3.5 Tree and Venn Diagrams

The last two displays of the chapter, for when a probability problem is complex enough to be worth drawing. A tree diagram lays out a sequence of stages with a probability or a frequency on every branch, so a path's probability is the product along it and a conditional probability is read by keeping one subtree and discarding the rest — which makes it the multiplication rule of section 3.3 drawn, with the conditional impossible to overlook. A Venn diagram shows two events as overlapping regions with a count in each, so unions, intersections and complements all become sums of regions. Closes by choosing among the tree, the Venn diagram and the contingency table.

Subject: Statistics · 69 slides · symbolic lesson

Open the interactive version of this deck

What this lesson covers

The lesson, slide by slide

1. Tree and Venn Diagrams

Title

Statistics · Chapter 3 — Probability Topics

Two pictures that do the probability rules for you: one for a sequence, one for an overlap

2. What this lesson gives you

Objectives

Sections 3.3 and 3.4 gave you rules and a table. This section gives you two pictures, and the pictures are worth learning because they make the rules nearly impossible to misuse. A tree draws an experiment that happens in stages; a Venn diagram draws two events that can overlap.

Keep one idea in view throughout. A tree is the multiplication rule drawn: every path is a product, and the second factor is always a conditional because it is drawn after the first branch. A Venn diagram is the addition rule drawn: the overlap sits in the middle, so you can see it and you cannot count it twice.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 189 — Tree and Venn Diagrams

3. Before anything new: one urn, one rule

Warm-up

Everything in this lesson is built from the multiplication rule of Section 3.3 and the idea of a conditional probability. Retrieve both before reading on.

Discussion prompt

An urn holds 3 red balls and 8 blue ones. You draw one ball and keep it out, then draw a second. Write the chance that the first is red. Then write the chance that the second is blue given that the first was red, and multiply the two. Which rule did you just use?

Notice what you had to keep track of in your head: which ball was gone, how many were left, which colour was now more likely. A tree diagram writes all of that down, one branch at a time, so that nothing has to be held in your head at all.

Notice also the answer, 24 out of 110. Hold on to it. It is one of the four leaves of the first probability tree you will draw, and you will see it again as a count in the frequency tree.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.3, p. 179 — the multiplication rule this lesson draws

4. Trees with frequencies

Section

Part 1

5. A tree lays out an experiment one stage at a time

Concept

A tree diagram is a graph for an experiment that happens in stages. Each stage is a column of branches. Each branch is one thing that can happen at that stage, and it is labelled with either how many ways it can happen, a frequency, or how likely it is, a probability.

tree diagram — A graph used to find the outcomes of an experiment. It consists of branches labelled with either frequencies or probabilities; the first set of branches is the first stage, the second set the second stage, and a path from the start to a leaf is one kind of outcome.

The book's first example is the urn you just met: 11 balls, 3 red and 8 blue, two draws. This time the first ball goes back before the second draw, which is what with replacement means. So both draws face the same 11 balls.

\[ 3 \text{ red} + 8 \text{ blue} = 11 \text{ balls, drawn twice, with replacement} \]

Read a tree left to right, like time. The first column of branches answers what happened on the first draw; from the end of each of those branches, a second column answers what happened on the second. A complete path answers both, so each path is one kind of outcome: RR, RB, BR or BB.

Order matters on a tree. Red then blue and blue then red are different paths, because the tree records which draw was which. That will turn out to be the most common thing students get wrong in this section.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 189 — Tree Diagrams, Example 3.24

6. Why there are 121 outcomes, not 4

Picture it

Four paths does not mean four equally likely outcomes. To see the real sample space, give every ball a name, as the book does: R1, R2, R3 for the reds and B1 to B8 for the blues. Now every ordered pair of names is one outcome.

Figure (svg): An 11 by 11 grid of small squares. Rows are the first ball drawn, labelled R1 to R3 then B1 to B8; columns are the second ball, labelled the same way. The top-left 3 by 3 block is red (RR), the top-right 3 by 8 block yellow (RB), the bottom-left 8 by 3 block orange (BR) and the bottom-right 8 by 8 block blue (BB). A legend gives the counts 9, 24, 24, 64 and the total 121.

Every square is one equally likely ordered pair. The four coloured blocks are the four leaves of the tree, and their areas are the leaf counts.

The grid has one row for each first ball and one column for each second ball, so it has 11 rows of 11 squares. Because the first ball is replaced, every one of the 121 squares is possible and every one is equally likely.

\[ 11 \times 11 = 121 \]

Now look at the colours. The red block, three rows by three columns, is every way to draw red then red: nine outcomes, R1R1 through R3R3. The yellow block is red then blue, three rows by eight columns. The four blocks are exactly the four paths of the tree.

\[ 9 + 24 + 24 + 64 = 121 \]

This picture is the reason a frequency tree works. A path's count is a rectangle's area, rows times columns, and the four rectangles tile the whole grid. That is why the counts are multiplied along a path and why the leaves always add back to the total.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 189 — Example 3.24: listing R1R1, R1R2, …

7. Multiply the counts along a path, add the leaves

Concept

Here is the grid redrawn as the book draws it, as a tree. The first branches carry 3 and 8, the numbers of ways the first draw can be red or blue. Each second-stage branch carries the same 3 and 8 again, because the urn was restored.

Figure (svg): A frequency tree. From the start, a branch labelled 3 R and a branch labelled 8 B. Each first-stage node splits again into 3 R and 8 B. The four leaves read RR: 3 × 3 = 9, RB: 3 × 8 = 24, BR: 8 × 3 = 24, BB: 8 × 8 = 64.

Multiplying the counts along a path gives the number of ordered outcomes on it; the four leaves always add back to the size of the sample space.

To get a leaf, multiply the numbers along its path. Three ways to pick a red first, and for each of those, eight ways to pick a blue second, gives 24 ordered outcomes on the red-then-blue path. That is the same three-rows-by-eight-columns block from the grid.

\[ \text{RB leaf} = 3 \times 8 = 24 \]

Then add every leaf. The total has to be the size of the whole sample space, because every outcome ends on exactly one leaf. This is your built-in check on the whole tree, and you should run it before answering anything.

\[ 9 + 24 + 24 + 64 = 121 = 11 \times 11 \]

Once the tree is checked, any probability for a single path is one leaf over the total. The next slide does exactly that for the book's questions.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 189 — Figure 3.2: Total = 64 + 24 + 24 + 9 = 121

8. Example 3.24(b)-(d) and (f): probabilities off the frequency tree

Worked example

Using the frequency tree, find the chance of two reds, of one of each colour in either order, of red first and blue second, and of two blues.

\[ \text{leaves: } RR = 9, \; RB = 24, \; BR = 24, \; BB = 64, \quad \text{total } 121 \]

Every outcome in the 121 is equally likely, so each probability is a count of outcomes over 121. The whole job is to decide which leaves an event uses, then add them.

(b) Two reds: one leaf

Why: only the RR path has red on both draws.

\[ P(RR) = \frac{9}{121} \approx 0.0744 \]

(b) The same answer as a product

Why: the book writes it as the chance of red, twice, since the draws are replaced.

\[ \left(\frac{3}{11}\right)\left(\frac{3}{11}\right) = \frac{9}{121} \]

(c) One of each colour: two leaves

Why: red then blue, or blue then red; distinct paths cannot happen together, so just add.

\[ P(RB\text{ OR }BR) = \frac{24}{121} + \frac{24}{121} \]

Add

Why: same denominator.

\[ P(RB\text{ OR }BR) = \frac{48}{121} \approx 0.3967 \]

(d) Red first AND blue second: one leaf

Why: this is the RB path only; blue then red has the colours the other way round.

\[ P(RB) = \frac{24}{121} \approx 0.1983 \]

(f) Two blues: one leaf

Why: the largest block of the grid.

\[ P(BB) = \frac{64}{121} \approx 0.5289 \]

In context: drawing twice with replacement from this urn, you get two blues a bit over half the time and one of each colour about 40 percent of the time.

Check that the four leaf probabilities add to 1

Why: one of each covers two leaves, so two reds, one of each, and two blues together are every outcome.

\[ \frac{9}{121} + \frac{48}{121} + \frac{64}{121} = \frac{121}{121} = 1 \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, pp. 189-190 — Example 3.24(b)-(d), (f)

9. Paths never overlap, so an OR of paths is a plain sum

Concept

In part (c) you added two leaves with no subtraction, and Section 3.3 should make you suspicious of that. The addition rule subtracts the overlap. Why was there nothing to subtract?

Because two different paths can never happen together. A pair of draws that went red then blue did not also go blue then red; the first draw was one colour or the other. So any two leaves are mutually exclusive, and their AND is empty.

\[ P(RB\text{ AND }BR) = 0 \]

With that zero in hand, the general addition rule loses its last term, exactly as it did for Klaus's vacations in Section 3.3.

\[ P(RB\text{ OR }BR) = P(RB) + P(BR) - 0 \]

This is one of the quiet advantages of a tree. It sorts every outcome into exactly one leaf, so an OR of leaves is always a plain sum and double counting is impossible. The careful work of the addition rule has been done by the drawing.

The warning runs the other way too. Events like red first and blue second are not single leaves; they are collections of leaves that can share one, here the RB leaf. For those, list the leaves each event uses and count the shared leaf only once.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 190 — Example 3.24(c)

10. How many leaves does each event use?

Sorting

Answering a tree question is mostly a matter of choosing leaves. Sort each event about two draws from the urn by the number of the four leaves it needs.

Sort into buckets

Sort each event by how many of the four leaves RR, RB, BR, BB it covers.

one leaf
both draws red; both draws blue
two leaves
one ball of each colour; red on the first draw; blue on the second draw
three leaves
at least one red; at most one blue
one
Both red is only RR and both blue is only BB. A single leaf means a single path, and its probability is that leaf over the total.
two
One of each is RB and BR. Red first is RR and RB, the whole upper subtree. Blue second is RB and BB, one leaf from each subtree.
three
At least one red is everything but BB, and at most one blue is everything but BB too. Three leaves are quicker as one minus the leaf left out.

Two patterns are worth noticing. An event about the first draw only, like red first, always takes a whole subtree. An event about the second draw only, like blue second, takes one leaf from each subtree, which is why it is harder to read off.

And whenever an event takes three of four leaves, count the one it leaves out instead. At least one red is 121 minus the 64 blue-blue outcomes, which is 57, so its probability is 57 out of 121, about 0.471.

\[ P(\text{at least one red}) = 1 - \frac{64}{121} = \frac{57}{121} \approx 0.471 \]

11. Try It 3.24: two face cards, with replacement

Worked example

A standard deck has 52 cards: 12 face cards, F, and 40 others, N. Draw two cards one at a time, with replacement. Find the probability that both are face cards.

Figure (svg): A frequency tree for two cards drawn with replacement. First branches 12 F and 40 N; each node splits into 12 F and 40 N. Leaves: FF 144, FN 480, NF 480, NN 1,600.

Same shape as the urn, bigger numbers. The FF leaf is the only one the question needs; the total is the denominator.

The deck plays the role of the urn: two kinds of object, drawn twice, the first put back. So the tree has exactly the same shape, with 12 and 40 on every branch in place of 3 and 8.

Count the FF leaf

Why: 12 face cards first, and for each, 12 face cards second.

\[ 12 \times 12 = 144 \]

Count the whole sample space

Why: 52 choices, then 52 again, since the card goes back.

\[ 52 \times 52 = 2704 \]

Divide

Why: one leaf over the total.

\[ P(FF) = \frac{144}{2704} \]

Simplify

Why: both numbers divide by 16.

\[ \frac{144}{2704} = \frac{9}{169} \approx 0.0533 \]

In context: with replacement, about 5 draws in every hundred pairs give two face cards.

Check the leaves and the product form

Why: the four leaves must total 2,704, and the probability must equal the chance of a face card squared, since the draws are replaced.

\[ 144 + 480 + 480 + 1600 = 2704, \qquad \left(\frac{12}{52}\right)^2 = \left(\frac{3}{13}\right)^2 = \frac{9}{169} \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 190 — Try It 3.24

12. Trap: counting RB and BR as one outcome

Trap

The trap

Asked for the chance of one ball of each colour, with replacement, a common first line:

\[ P(\text{one of each}) = \frac{24}{121} \]

\[ \approx 0.198 \]

Wrong. It uses only the red-then-blue leaf.

It feels right because one of each sounds like one outcome. On the tree it is not: the tree records order, and there are two paths that give one of each.

The fix

One of each happens on the red-then-blue path or on the blue-then-red path, and the two are different outcomes:

\[ \frac{24}{121} + \frac{24}{121} = \frac{48}{121} \]

\[ \approx 0.397 \]

The habit that prevents this: before adding, name every leaf the event uses. If the event does not say which draw comes first, it almost always takes both orders.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 190 — Example 3.24(c)

13. Trees with probabilities

Section

Part 2

14. From counts to probabilities: divide every branch

Concept

A frequency tree needs the whole sample space counted, and that gets unwieldy fast. The fix is to label each branch with a probability instead: the fraction of the time that branch is taken, given where you are in the tree.

Start from the frequency tree and divide each branch count by the number of balls it chose from. The first red branch, 3 out of 11 balls, becomes the probability 3 out of 11. With replacement, every second-stage branch is again out of 11.

\[ 3 \longrightarrow \frac{3}{11}, \qquad 8 \longrightarrow \frac{8}{11} \]

Then a leaf becomes a product of probabilities instead of a product of counts. The red-then-red leaf is 3 over 11 times 3 over 11, and that is exactly the count 9 divided by the total 121: dividing each factor by 11 divided the product by 121.

\[ \frac{3}{11} \cdot \frac{3}{11} = \frac{9}{121} \]

So a probability tree carries the same information as a frequency tree, already divided through. The numbers at the ends of the branches are found by multiplying the numbers on the branches of the path, and the leaves now add to 1 instead of to the size of the sample space.

path probability — The probability of one complete path through a tree: the product of the probabilities on its branches. The paths are mutually exclusive and cover every outcome, so the path probabilities add to 1.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 190 — Example 3.25: branches labelled with probabilities

15. Reading a path probability piece by piece

Notation

Every path probability is the multiplication rule of Section 3.3, written along the branches. Here is the red-then-blue path of the without-replacement urn, decoded.

Annotate

On: \( P(R\text{ AND }B) = P(R)\;P(B \mid R) = \frac{3}{11} \cdot \frac{8}{10} \)

  • The leaf: red on the first draw and blue on the second. Order is part of the event; blue then red is a different leaf.
  • The first branch. Nothing has happened yet, so it is a plain probability: 3 reds among 11 balls.
  • The second branch. It hangs off the red branch, so it is measured only in the world where the first ball was red. The book writes it P(B | R), short for blue on the second given red on the first.
  • After a red is removed, 10 balls remain and all 8 blues are still there. The denominator dropped because the draw was without replacement.

The key idea is where the second branch is drawn. Because it starts at the end of the first branch, it cannot help being a conditional. You never have to remember to condition; the tree conditions for you.

That is also why the tree is such a good cure for the error in Section 3.3 of multiplying two plain probabilities. On a tree, the second number physically sits after the first, and you fill it in by asking what is left in the urn now.

The shorthand needs care. When the book writes the probability of B given R on a tree, both letters refer to different draws: the B is about the second draw and the R about the first. Say it in full at least once for every tree you draw.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, pp. 191-192 — the general labelling of a probability tree

16. The general tree: every second branch is a conditional

Picture it

Strip the numbers off and the tree shows its structure. This is the book's general labelling for any two-draw problem with red and blue.

Figure (svg): A probability tree with symbolic labels. First branches P(R) and P(B). After R: P(R | R) and P(B | R). After B: P(R | B) and P(B | B). The leaves read P(R AND R), P(R AND B), P(B AND R), P(B AND B), each written as the product of its two branches.

Every second-stage label is a conditional: it is measured only among the outcomes that took the branch before it.

The first column holds two plain probabilities, and they add to 1 because the first draw is one colour or the other. Each pair in the second column holds two conditionals measured in the same world, so each pair adds to 1 as well.

\[ P(R) + P(B) = 1, \qquad P(R \mid R) + P(B \mid R) = 1 \]

Now look at the leaves. Each one is the multiplication rule: the first branch times the conditional branch after it. The tree does not need a separate rule for AND; the product along the path is the AND.

The last thing the picture tells you is where independence lives. If the two second-stage pairs carry identical labels, the second draw does not care what the first was, and the draws are independent. With replacement they are identical; without replacement they differ. You can see independence by comparing the two subtrees.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 192 — P(R | R) here means P(R on 2nd | R on 1st)

17. Without replacement: more or fewer red pairs?

Prediction

Now draw the two balls without replacement: the first red or blue ball stays out. Before building the new tree, predict how one number moves.

Predict first

With replacement, the chance of two reds was 9 out of 121, about 0.074. Without replacement, the chance of two reds is

  • smaller than 0.074
  • larger than 0.074
  • exactly 0.074
  • impossible to say without the tree

Correct: smaller than 0.074

Why: After a red is removed, only 2 of the remaining 10 balls are red, so the second red becomes less likely than 3 in 11. The chance of two reds is 3 over 11 times 2 over 10, which is 6 over 110, about 0.055.

The reasoning takes one sentence: taking a red out makes red scarcer for the second draw. You do not need the whole tree to predict the direction, only the second branch on the red-red path.

The same reasoning predicts that two blues also become less likely, since taking a blue out makes blue scarcer. Then something else must become more likely, because the four paths always total 1. Keep that in mind; the next slides confirm it.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 190 — Example 3.25: this time without replacement

18. Example 3.25: building the tree without replacement

Worked example

An urn has 3 red and 8 blue marbles. Draw two, one at a time, without replacement. Build the probability tree and find all four path probabilities.

Work stage by stage and, at every node, ask one question: what is in the urn right now? The answer to that question is the whole method.

First stage

Why: all 11 marbles are in the urn.

\[ P(R) = \frac{3}{11}, \qquad P(B) = \frac{8}{11} \]

Second stage after a red

Why: a red is out, leaving 2 red and 8 blue among 10.

\[ P(R \mid R) = \frac{2}{10}, \qquad P(B \mid R) = \frac{8}{10} \]

Second stage after a blue

Why: a blue is out, leaving 3 red and 7 blue among 10.

\[ P(R \mid B) = \frac{3}{10}, \qquad P(B \mid B) = \frac{7}{10} \]

Multiply along the two red-first paths

Why: first branch times second branch.

\[ P(RR) = \frac{3}{11} \cdot \frac{2}{10} = \frac{6}{110}, \quad P(RB) = \frac{3}{11} \cdot \frac{8}{10} = \frac{24}{110} \]

Multiply along the two blue-first paths

Why: same move, lower subtree.

\[ P(BR) = \frac{8}{11} \cdot \frac{3}{10} = \frac{24}{110}, \quad P(BB) = \frac{8}{11} \cdot \frac{7}{10} = \frac{56}{110} \]

Figure (svg): A probability tree for two draws without replacement from 3 red and 8 blue. First branches R: 3/11 and B: 8/11. After R: R 2/10 and B 8/10. After B: R 3/10 and B 7/10. Leaves: RR 6/110, RB 24/110, BR 24/110, BB 56/110.

The second-stage labels differ between the two subtrees: a red already taken leaves only 2 reds among 10. That difference is what without replacement means.

Keep the leaves over 110 rather than simplifying. With a common denominator, adding leaves is just adding numerators, and every answer in the next slide becomes a one-line sum.

Check that the four paths total 1

Why: every pair of draws follows exactly one path.

\[ \frac{6 + 24 + 24 + 56}{110} = \frac{110}{110} = 1 \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, pp. 190-191 — Example 3.25, Figure 3.4

19. Example 3.25(a), (b), (d), (e): reading the tree

Worked example

Using the tree you just built, find the chance of two reds, of one of each colour, of red first and blue second, and of two blues.

\[ RR = \frac{6}{110}, \; RB = \frac{24}{110}, \; BR = \frac{24}{110}, \; BB = \frac{56}{110} \]

Each question names leaves, and each leaf is already computed, so every part is a lookup or a sum. The only thinking is choosing the leaves.

(a) Two reds: one leaf

Why: the product 3 over 11 times 2 over 10.

\[ P(RR) = \frac{6}{110} \approx 0.0545 \]

(b) One of each: two leaves

Why: the book's fill-in-the-blanks line; the missing factors are the blue-then-red path.

\[ P(RB\text{ OR }BR) = \left(\frac{3}{11}\right)\left(\frac{8}{10}\right) + \left(\frac{8}{11}\right)\left(\frac{3}{10}\right) \]

Add the two leaves

Why: twenty-four plus twenty-four, over 110.

\[ P(RB\text{ OR }BR) = \frac{48}{110} \approx 0.4364 \]

(d) Red first AND blue second

Why: a single leaf, the upper mixed one.

\[ P(RB) = \left(\frac{3}{11}\right)\left(\frac{8}{10}\right) = \frac{24}{110} \approx 0.2182 \]

(e) Two blues

Why: the lowest leaf.

\[ P(BB) = \left(\frac{8}{11}\right)\left(\frac{7}{10}\right) = \frac{56}{110} \approx 0.5091 \]

In context: without replacement you get one of each colour about 44 percent of the time and two blues about 51 percent of the time.

Check against the prediction

Why: two reds fell from 0.0744 with replacement to 0.0545, as predicted, and one of each rose from 0.3967 to 0.4364.

\[ \frac{6}{110} < \frac{9}{121}, \qquad \frac{48}{110} > \frac{48}{121} \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 191 — Example 3.25(a), (b), (d), (e)

20. Example 3.25(b): fill in the blanks

Fill the middle

The book prints part (b) with two factors missing. Fill them in from the tree: the second product is the blue-then-red path.

Fill in the blanks

P(RB\text3BR) = \left(\frac48___\right)\left(\frac______\right) + \left(\frac______\right)\left(\frac___}___\right) = \frac___}___

Why: After a blue is taken first, all 3 reds are still in the urn and 10 marbles remain, so the second branch is 3 out of 10. Eight times three is 24, and 24 plus 24 is 48, over 110.

The blank that trips people is the first one. After a blue draw the reds are untouched, so the numerator stays 3; only the denominator falls to 10. Writing 2 there would be taking a red out when a blue was drawn.

Look at the two products once they are complete. They have the same numerators in a different order, 3 times 8 and 8 times 3, which is why the two mixed leaves are equal. That symmetry holds for any two draws without replacement, and it is a quick check on your tree.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 191 — Example 3.25(b)

21. With and without replacement, side by side

Comparison

Fill the blanks from the two trees, then read down the last column. It says which paths replacement helps and which it hurts.

Comparison matrix

pathwith replacementwithout replacementwithout is
RR9/121 ≈ 0.0746/110smaller
RB24/121 ≈ 0.19824/110 ≈ 0.218larger
BR24/12124/110 ≈ 0.218larger
BB64/121 ≈ 0.52956/110 ≈ 0.509smaller

Every same-colour path loses and every mixed path gains. That is the pattern you predicted, and it is general: removing a ball of one colour makes that colour scarcer and the other colour more common on the next draw.

Figure (svg): A grouped bar chart of the four path probabilities. For each path, a teal bar for with replacement and an orange bar for without: RR 0.074 and 0.055, RB 0.198 and 0.218, BR 0.198 and 0.218, BB 0.529 and 0.509.

Without replacement, the same-colour paths shrink and the mixed paths grow. The total is 1 either way, so what one pair loses the other gains.

Both columns still add to 1. Replacement does not create or destroy probability; it moves some from the same-colour paths to the mixed ones. That is why the grouped bars rise in the middle and fall at the ends.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, pp. 189-191 — Examples 3.24 and 3.25

22. Why the mixed paths gain when nothing goes back

Intuition

The table shows a pattern; here is the reason for it. Think about what the first draw does to the urn in each case.

If the first ball is red, the urn now holds a smaller share of red than before: 2 in 10 against 3 in 11. So a second red is less likely and a second blue more likely. Removing a colour always pushes the next draw toward the other colour.

\[ \frac{2}{10} = 0.200 < 0.273 \approx \frac{3}{11}, \qquad \frac{8}{10} = 0.800 > 0.727 \approx \frac{8}{11} \]

The same happens after a blue: 7 blue in 10 is a smaller share than 8 in 11, so a second blue is less likely and a red more likely. Either way the second draw leans away from the first, and leaning away means switching colour.

\[ \frac{7}{10} = 0.700 < 0.727 \approx \frac{8}{11}, \qquad \frac{3}{10} = 0.300 > 0.273 \approx \frac{3}{11} \]

That is the whole story behind the bar chart: sampling without replacement makes matching pairs rarer and mixed pairs commoner. It is a real effect in card games, in lotteries and in quality inspections, and it shrinks as the urn gets bigger, which a later slide lets you see.

It also tells you what dependence means here. The draws are dependent because the second draw's chances change with the first, and they change in a predictable direction: toward whatever colour is now relatively more common.

23. Trap: keeping the second stage unchanged without replacement

Trap

The trap

Asked for two reds when the first marble is not put back:

\[ P(RR) = \frac{3}{11} \cdot \frac{3}{11} \]

\[ = \frac{9}{121} \approx 0.074 \]

Wrong. That is the with-replacement answer.

It is easy to write because both factors are the chance of red, and the chance of red was 3 out of 11 a moment ago. But the urn has changed.

The fix

The second branch hangs off a red draw, so it is measured in an urn with 2 red among 10:

\[ P(RR) = \frac{3}{11} \cdot \frac{2}{10} \]

\[ = \frac{6}{110} \approx 0.055 \]

The habit that prevents this: before writing any second-stage label, say what is in the urn now, out loud. Without replacement, the denominator always drops by one, and the numerator drops by one on the same-colour branch.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 191 — the Note after Figure 3.4

24. At least and at most: count the leaf you leave out

Concept

Try It 3.25 asks for at most one face card and at least one face card, and the book adds a hint for each. The hints translate the words into leaves, and that translation is the whole difficulty.

At most one means zero or one: no face card, or exactly one. On a two-card tree that is the NN leaf plus the two mixed leaves, FN and NF. The only leaf excluded is FF, where there are two face cards.

\[ \text{at most one F} = NN\text{ OR }FN\text{ OR }NF = \text{everything except } FF \]

At least one means one or two: the two mixed leaves plus FF. The only leaf excluded is NN, where there are no face cards.

\[ \text{at least one F} = FN\text{ OR }NF\text{ OR }FF = \text{everything except } NN \]

Both events use three of the four leaves, so both are quickest by the complement rule: one minus the single leaf left out. That is one subtraction instead of adding three leaves, and it is harder to get wrong.

A way to remember which leaf to drop: at most one forbids the top count, two face cards; at least one forbids the bottom count, zero. Draw the tree, cross out the forbidden leaf, and whatever is left is your event.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 192 — Try It 3.25(c) and (d), with the book's hints

25. Try It 3.25: face cards without replacement

Worked example

Draw two cards from a 52-card deck without replacement: 12 face cards, F, and 40 others, N. Find the chance of one of each, of a non-face card given a face card first, of at most one face card, and of at least one face card.

Figure (svg): A probability tree for two cards drawn without replacement. First branches F: 12/52 and N: 40/52. After F: F 11/51 and N 40/51. After N: F 12/51 and N 39/51. Leaves: FF 132/2652, FN 480/2652, NF 480/2652, NN 1560/2652. The FN and NF leaves are highlighted as one of each.

At least one face card is every leaf but NN; at most one is every leaf but FF. Each is quickest as one minus a single leaf.

The book's tree already has every label. Work in counts over 2,652, the product 52 times 51: the leaves are then 132, 480, 480 and 1,560, which add to 2,652.

(a) One of each: two leaves

Why: face then non-face, or non-face then face.

\[ P(FN\text{ OR }NF) = \frac{480 + 480}{2652} = \frac{960}{2652} \approx 0.362 \]

(b) Non-face given face first: read the branch

Why: the conditional is the label on the branch after F.

\[ P(N \mid F) = \frac{40}{51} \approx 0.784 \]

(c) At most one face card: all but FF

Why: zero or one face card leaves out only the two-face leaf.

\[ P(\text{at most one F}) = 1 - \frac{132}{2652} = \frac{2520}{2652} \approx 0.950 \]

(d) At least one face card: all but NN

Why: one or two face cards leaves out only the no-face leaf.

\[ P(\text{at least one F}) = 1 - \frac{1560}{2652} = \frac{1092}{2652} \approx 0.412 \]

In context: in two cards dealt from a fresh deck, you get at least one face card about 41 percent of the time, and two face cards only about 5 percent of the time.

Check part (d) by adding the three leaves directly

Why: FF, FN and NF are the leaves with a face card; their sum must match the complement.

\[ 132 + 480 + 480 = 1092 \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 192 — Try It 3.25, Figure 3.5

26. Example 3.26(a): which product is two tabbies?

Elimination

A litter has 4 tabby kittens and 5 black ones. A family chooses two at random, without replacement. The book offers four products for the chance that both are tabby. Knock out the wrong ones.

Eliminate the wrong options

Which product gives the probability that both kittens are tabby?

  • k. (4/9)(3/8)
  • h. (1/2)(1/2)
  • w. (4/9)(4/9)
  • m. (4/9)(5/9)

Survives elimination: k

Why: The first tabby has probability 4 out of 9. Given that, 3 tabbies remain among 8 kittens, so the second branch is 3 out of 8. The product is 12 over 72, which is one sixth.

Try the elimination by asking two questions of each product. Is the first factor the share of tabbies among all nine kittens? Is the second factor the share of tabbies among the eight left after a tabby went home? Only one product answers yes twice.

Each wrong product is a different way to skip a question. One half skips counting, four ninths twice skips asking whether the kitten came back, and four ninths times five ninths skips checking which colour the second branch is.

\[ P(TT) = \frac{4}{9} \cdot \frac{3}{8} = \frac{12}{72} = \frac{1}{6} \approx 0.1667 \]

The book's multiple choice is a useful model for your own checking. Before trusting a product, read each factor aloud with its condition: tabby first, out of all nine; tabby second, given a tabby is gone, out of eight.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, pp. 192-193 — Example 3.26(a)

27. Example 3.26(b)-(d): the rest of the kittens

Worked example

Same litter, two kittens chosen without replacement. Find the chance of one of each colour, of a tabby second given a black first, and of two kittens of the same colour.

Figure (svg): A probability tree for two kittens chosen without replacement from 4 tabby and 5 black. First branches T: 4/9 and B: 5/9. After T: T 3/8 and B 5/8. After B: T 4/8 and B 4/8. Leaves TT 12/72, TB 20/72, BT 20/72, BB 20/72.

The thick branch is part (c): a tabby second, given a black first, is a branch label and needs no arithmetic at all.

Put every leaf over 72, the product 9 times 8, and the tree reads 12, 20, 20 and 20. Checking: those add to 72. Every part is now a choice of leaves.

(b) One of each: two leaves

Why: tabby then black, or black then tabby; this is the book's choice (d).

\[ P(TB\text{ OR }BT) = \left(\frac{4}{9}\right)\left(\frac{5}{8}\right) + \left(\frac{5}{9}\right)\left(\frac{4}{8}\right) \]

Evaluate

Why: twenty over 72 plus twenty over 72.

\[ P(TB\text{ OR }BT) = \frac{40}{72} \approx 0.556 \]

(c) Tabby second given black first: a branch

Why: the thick branch after B; all 4 tabbies remain among 8.

\[ P(T\text{ on 2nd} \mid B\text{ on 1st}) = \frac{4}{8} = 0.5 \]

(d) Same colour: two leaves

Why: two tabbies or two blacks.

\[ P(TT\text{ OR }BB) = \frac{12}{72} + \frac{20}{72} = \frac{32}{72} \approx 0.444 \]

In context: a family choosing two kittens at random from this litter goes home with a matching pair a bit under half the time.

Check parts (b) and (d) against each other

Why: a pair is either one of each or the same colour, never both, so the two answers must add to 1.

\[ \frac{40}{72} + \frac{32}{72} = \frac{72}{72} = 1 \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 193 — Example 3.26(b)-(d); the book leaves (d) as 32/72

28. Try It 3.26: one red and one yellow

Worked example

A box holds 4 red balls and 3 yellow ones. Two are drawn without replacement. Find the probability that one ball of each colour is selected.

You do not need to draw the whole tree to answer, but you do need the two paths it would show. One of each means red then yellow, or yellow then red.

Red then yellow

Why: 4 red of 7, then 3 yellow of the 6 left.

\[ P(RY) = \frac{4}{7} \cdot \frac{3}{6} = \frac{12}{42} \]

Yellow then red

Why: 3 yellow of 7, then all 4 red of the 6 left.

\[ P(YR) = \frac{3}{7} \cdot \frac{4}{6} = \frac{12}{42} \]

Add the two paths

Why: distinct paths never overlap.

\[ P(RY\text{ OR }YR) = \frac{24}{42} \]

Simplify

Why: divide top and bottom by 6.

\[ \frac{24}{42} = \frac{4}{7} \approx 0.571 \]

Look at the two products before adding them. The numerators are 4 times 3 and 3 times 4, and the denominators are 7 times 6 both times. The two orders are equally likely, which is the same symmetry you saw in the urn: the mixed paths always match.

In context: more often than not, the two balls differ in colour. That is typical when the two colours are nearly balanced, because a mixed pair can happen in two orders.

Check with the complement

Why: same colour is two reds or two yellows; one minus that must give the same answer.

\[ 1 - \left(\frac{4}{7} \cdot \frac{3}{6} + \frac{3}{7} \cdot \frac{2}{6}\right) = 1 - \frac{18}{42} = \frac{24}{42} \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 193 — Try It 3.26

29. Exercise 116: Miguel eats two cookies

Worked example

A box has 3 chocolate cookies and 7 butter cookies. Miguel picks one at random and eats it, then picks another and eats it. Build the tree and find the chance of the same flavour twice, of different flavours, and of butter on the second cookie.

The book's aside, how many cookies did he take, is the whole clue. An eaten cookie cannot go back in the box, so this is without replacement, and the second-stage labels are out of 9.

Figure (svg): A probability tree for two cookies eaten without replacement from 3 chocolate and 7 butter. First branches C: 3/10 and B: 7/10. After C: C 2/9 and B 7/9. After B: C 3/9 and B 6/9. Leaves CC 6/90, CB 21/90, BC 21/90, BB 42/90.

Two leaves are same flavour and two are different, and together they are everything, so the two answers must add to 1.

The four paths

Why: multiply along each; the numerators are 3 times 2, 3 times 7, 7 times 3, and 7 times 6.

\[ CC = \frac{6}{90}, \; CB = \frac{21}{90}, \; BC = \frac{21}{90}, \; BB = \frac{42}{90} \]

(b) Is the second flavour independent of the first?

Why: no: the chance of chocolate second is 2 out of 9 after a chocolate but 3 out of 9 after a butter.

\[ P(C \mid C) = \frac{2}{9} \ne \frac{3}{9} = P(C \mid B) \]

(d) Same flavour

Why: the two same leaves.

\[ P(S) = \frac{6}{90} + \frac{42}{90} = \frac{48}{90} \approx 0.533 \]

(e) Different flavours, by the branches

Why: the two mixed leaves.

\[ P(T) = \frac{21}{90} + \frac{21}{90} = \frac{42}{90} \approx 0.467 \]

(f) Butter second

Why: one leaf from each subtree.

\[ P(U) = \frac{21}{90} + \frac{42}{90} = \frac{63}{90} = 0.7 \]

In context: Miguel's second cookie is butter 70 percent of the time, exactly the share of butter cookies in the box. That is not a coincidence; Part 3 explains it.

Check part (e) by the complement rule, as the book asks

Why: different flavours is the complement of same flavour, so both methods must agree.

\[ P(T) = 1 - P(S) = 1 - \frac{48}{90} = \frac{42}{90} \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5 homework, p. 215 — Exercise 116

30. Exercises 114 and 115: a coin, then a bead

Worked example

An unfair coin lands heads with probability 2 out of 3. Then one bead is drawn from a cup of 3 red, 4 yellow and 5 blue. Find the chance of heads and a red bead, and the chance of a blue bead.

Figure (svg): A tree with an unfair coin first, H 2/3 and T 1/3, then a bead from 3 red, 4 yellow and 5 blue: each node splits into R 3/12, Y 4/12 and B 5/12. Six leaves: H AND R 6/36, H AND Y 8/36, H AND B 10/36, T AND R 3/36, T AND Y 4/36, T AND B 5/36. The two blue leaves are highlighted.

Blue can arrive by two routes, so its probability is the sum of two leaves from different subtrees.

Two things are new here. The second stage has three branches, not two, so the tree has six leaves. And the cup does not depend on the coin, so both subtrees carry the same labels: the coin and the bead are independent.

114: Heads AND red, one path

Why: multiply the coin branch by the bead branch.

\[ P(H\text{ AND }R) = \frac{2}{3} \cdot \frac{3}{12} = \frac{6}{36} \]

Simplify

Why: the book's choice (c) leaves it as 6 over 36.

\[ \frac{6}{36} = \frac{1}{6} \approx 0.1667 \]

115: Blue, two paths

Why: blue can follow heads or tails, so add one leaf from each subtree.

\[ P(\text{Blue}) = \frac{2}{3} \cdot \frac{5}{12} + \frac{1}{3} \cdot \frac{5}{12} \]

Evaluate

Why: ten over 36 plus five over 36; the book's choice (a).

\[ P(\text{Blue}) = \frac{10}{36} + \frac{5}{36} = \frac{15}{36} \approx 0.417 \]

Blue is an event about the second stage alone, so it takes one leaf from each subtree, just like red second in the urn. The coin's two branches weight the two blue leaves, two thirds and one third, but because the blue label is 5 out of 12 on both, the weights add back to 1.

In context: a blue bead turns up about 42 percent of the time, and the coin has nothing to do with it.

Check using independence

Why: since the cup ignores the coin, the chance of blue must be just the share of blue beads in the cup.

\[ \frac{15}{36} = \frac{5}{12} \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5 homework, p. 215 — Exercises 114-115, Figure 3.14

31. A big enough urn forgets the draw

Tweak it

Without replacement changes the second draw because the urn is small. Here is the chance of two reds without replacement, for an urn with a fixed proportion p of red balls and N balls in all. Across the bottom is N; up the side is the chance of two reds.

Parameter explorer

Set p to 0.27, close to the book's 3 red in 11. Read the height at N = 11, then at N = 110. Compare both with p squared, about 0.074, the with-replacement answer.

\[ P(RR) = {p} \cdot \frac{{p} N - 1}{N - 1} \]

  • p — from 0.1 to 0.33: proportion red, p

The formula is just the red-red path: the first branch is the proportion p, and the second is one fewer red over one fewer ball. At N equal to 11 with 3 reds, it gives 6 over 110, the book's answer.

\[ P(RR) = \frac{3}{11} \cdot \frac{3 - 1}{11 - 1} = \frac{6}{110} \approx 0.0545 \]

As N grows the curve climbs toward a level line at p squared. For the book's proportion it is about 0.065 at N of 22, 0.071 at 55, and 0.0726 at 110, against a ceiling of 0.0744. Removing one ball from a huge urn barely changes it.

This is the reason later chapters can treat a sample drawn without replacement from a large population as if the draws were independent. The tree is still the right picture; its two subtrees just carry nearly identical labels.

32. Conditionals from a tree

Section

Part 3

33. Conditioning on the first draw: keep one subtree

Concept

A conditional probability restricts the sample space to the outcomes where the condition holds. On a tree, when the condition is about the first stage, that restriction is easy to see: you keep one subtree and throw the other away.

Figure (svg): The frequency tree for Example 3.24 with the red-first subtree faded out. The blue-first branch and its two children are bright: BR 24, highlighted, and BB 64. A note reads: given blue first, only 24 + 64 = 88 outcomes remain.

Conditioning on the first draw throws away a whole subtree. What remains is a smaller sample space of 88, and 24 of those are red second.

Take the book's question: red on the second draw, given blue on the first, with replacement. The condition blue first discards the whole red-first subtree. What remains are the 24 blue-then-red outcomes and the 64 blue-then-blue ones.

\[ \text{reduced sample space} = 24 + 64 = 88 \]

Among those 88, red on the second draw means the BR leaf, 24 outcomes. The conditional probability is the wanted count over the reduced count, not over 121.

\[ P(R\text{ on 2nd} \mid B\text{ on 1st}) = \frac{24}{88} \]

This is the definition of a conditional from Section 3.1, done with the picture instead of the formula. The tree makes the reduced sample space something you can point at, which is the step that the formula hides.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 190 — Example 3.24(e): the sample space has been reduced

34. Reading a conditional as a fraction of leaves

Notation

Every conditional you read from a tree has the same shape: some leaves on top, more leaves on the bottom, and the top leaves always among the bottom ones.

Annotate

On: \( P(R\text{ on 2nd} \mid B\text{ on 1st}) = \frac{\text{BR}}{\text{BR} + \text{BB}} = \frac{24}{88} \)

  • Everything to the right of the bar is the condition: here, blue on the first draw. It decides which leaves survive.
  • The surviving leaves in which the event on the left also happens: blue first and red second. It is the AND of the two events.
  • All surviving leaves: the reduced sample space. It is the probability, or count, of the condition alone.
  • A count over a count because this is a frequency tree. On a probability tree you divide leaf probabilities the same way.

Written as a formula this is the familiar definition: the probability of the AND over the probability of the condition. The tree simply tells you which leaves make up each piece.

\[ P(A \mid B) = \frac{P(A\text{ AND }B)}{P(B)} \]

On a probability tree the leaves are fractions of 1 rather than counts, but the recipe is identical. For the same question you would divide 24 over 121 by 88 over 121, and the two 121s cancel to leave 24 over 88. Counts and probabilities give the same conditional because a conditional is a ratio.

One test protects you from most mistakes: the top leaves must be a subset of the bottom leaves. If a leaf appears on top but not below, you have divided by the wrong event.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 190 — Example 3.24(e) and (g)

35. Example 3.24(e) and (g): two conditionals, with replacement

Worked example

Using the frequency tree with replacement, find the chance of red on the second draw given blue on the first, and of blue on the second given red on the first.

\[ RR = 9, \; RB = 24, \; BR = 24, \; BB = 64 \]

Both conditions are about the first draw, so both answers come from keeping one subtree. Write down the reduced sample space before the answer; it is the step that makes the conditional a conditional.

(e) Keep the blue-first subtree

Why: BR and BB survive.

\[ 24 + 64 = 88 \]

Wanted leaf over surviving leaves

Why: red second is the BR leaf.

\[ P(R\text{ on 2nd} \mid B\text{ on 1st}) = \frac{24}{88} \]

Simplify

Why: divide top and bottom by 8.

\[ \frac{24}{88} = \frac{3}{11} \approx 0.273 \]

(g) Keep the red-first subtree

Why: RR and RB survive.

\[ 9 + 24 = 33 \]

Wanted leaf over surviving leaves, simplified

Why: blue second is the RB leaf; divide top and bottom by 3.

\[ P(B\text{ on 2nd} \mid R\text{ on 1st}) = \frac{24}{33} = \frac{8}{11} \approx 0.727 \]

In context: after a blue first draw, a red second draw is still 3 chances in 11, exactly as likely as it was before. The replaced ball has left the urn as it was.

Check against the branch labels

Why: with replacement, the branch after blue is labelled 3 out of 11 and the branch after red for blue is 8 out of 11; the leaf arithmetic reproduces both.

\[ \frac{24}{88} = \frac{3}{11}, \qquad \frac{24}{33} = \frac{8}{11} \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 190 — Example 3.24(e), (g)

36. Forward conditionals are already on the branches

Concept

That check points at a shortcut. When the condition is the first draw and the event is the second draw, the conditional you want is exactly the label on a second-stage branch. You built it in when you drew the tree.

Example 3.25 asks two such questions for the tree without replacement. Both are read straight off the branches, with no leaves and no division.

\[ P(R\text{ on 2nd} \mid B\text{ on 1st}) = \frac{3}{10}, \qquad P(B\text{ on 2nd} \mid R\text{ on 1st}) = \frac{8}{10} \]

You can confirm the first one the long way. The blue-first subtree holds 24 over 110 and 56 over 110, so it has 80 over 110 in all, and the blue-then-red leaf is 24 of that. Dividing gives the branch label back.

\[ \frac{24}{110} \div \frac{80}{110} = \frac{24}{80} = \frac{3}{10} \]

Call these forward conditionals: they run in the same direction as the tree, from an earlier stage to a later one. They are free. The ones that take work are the backward conditionals, where the condition is a later stage. That is the next slide but one.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 191 — Example 3.25(c) and (f)

37. Example 3.25(f): the book's shorthand slips

Error analysis

The book's printed answer to part (f) reads as follows. The number is right. One piece of the notation is not.

Annotate

On: \( P(B\text{ on 2nd} \mid R\text{ on 1st}) = P(R \mid B) = \frac{8}{10} \)

  • The question: blue on the second draw, given red on the first. The condition, to the right of the bar, is the red first draw.
  • A misprint. In the book's own shorthand this means red on the 2nd given blue on the 1st, which is 3 out of 10. The shorthand for part (f) is P(B | R).
  • Correct for the question asked: after a red is removed, 8 blues remain among 10 marbles.

The slip matters because the two conditionals are different numbers on this tree: blue after red is 8 out of 10, red after blue is 3 out of 10. Swapping the letters around the bar swaps which subtree you are standing in.

\[ P(B \mid R) = \frac{8}{10} \ne \frac{3}{10} = P(R \mid B) \]

The protection is the book's own list on page 192: in tree shorthand, the letter to the right of the bar is always the first draw. Read the shorthand back in full, second given first, and a swapped pair of letters shows at once.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, pp. 191-192 — Example 3.25(f) and the labelling list

38. Is the second draw less likely to be red?

Anomaly

Without replacement, the second draw's chances change after every first draw: 2 out of 10 after a red, 3 out of 10 after a blue. So what happens to the plain chance that the second ball is red, before you know anything about the first?

Predict first

Without replacement, the probability that the second ball is red, not knowing the first, is

  • exactly 3/11, the same as the first draw
  • 2/10, since a ball is gone
  • 3/10
  • somewhere between 2/10 and 3/10 but not 3/11

Correct: exactly 3/11, the same as the first draw

Why: Red second collects the red-red leaf and the blue-red leaf: 6 over 110 plus 24 over 110 is 30 over 110, which is exactly 3 over 11. Averaged over what the first draw might have been, the second draw sees the urn just as the first did.

\[ P(R\text{ on 2nd}) = \frac{6}{110} + \frac{24}{110} = \frac{30}{110} = \frac{3}{11} \]

It feels as if removing a ball should change things, and it does, but only once you know which ball. Before you know, the second ball is just as likely as the first to be any particular ball in the urn, so its chance of being red is the share of red in the urn.

You met this already: Miguel's second cookie was butter 70 percent of the time, the butter share of the box. Without replacement makes the draws dependent, not the second draw different on its own. Both statements are true at once.

39. Conditioning on the second draw: collect leaves across the tree

Concept

Now reverse the question. Given that the second marble is blue, what is the chance that the first was red? The condition is about the later stage, so it does not pick out a subtree. It picks out one leaf from each.

Figure (svg): The without-replacement tree of Example 3.25 with the two leaves that end in blue highlighted: RB 24/110 and BB 56/110, one in each subtree. RR and BR are faded. A note reads B on 2nd: 24/110 + 56/110 = 80/110.

Conditioning on the second draw keeps leaves from both subtrees. The new sample space is spread across the tree, which is why the answer is not on any branch.

The condition: blue on the second draw

Why: the RB leaf and the BB leaf, one in each subtree.

\[ P(B\text{ on 2nd}) = \frac{24}{110} + \frac{56}{110} = \frac{80}{110} \]

The AND: red first and blue second

Why: just the RB leaf.

\[ P(R\text{ on 1st}\text{ AND }B\text{ on 2nd}) = \frac{24}{110} \]

Divide the AND by the condition

Why: the 110s cancel.

\[ P(R\text{ on 1st} \mid B\text{ on 2nd}) = \frac{24}{80} = 0.3 \]

This number is on no branch. The tree runs from first draw to second, and you asked a question that runs from second to first, so you had to collect leaves and divide. Every backward conditional works this way, and it is the one kind of tree question that is never free.

Compare the answer with the plain chance of red first, 3 out of 11, about 0.273. Learning the second ball was blue raised it slightly to 0.3: a blue second is a little more likely when the first ball took a red out, leaving blue more common.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 191 — the Example 3.25 tree, read backwards

40. Free off a branch, or work across the tree?

Discrimination

Before computing anything from a tree, decide what kind of question you have. Sort each question about the without-replacement urn.

Sort into buckets

Sort each question by how you get it from the tree.

read one branch label
P(R on 2nd | B on 1st); P(B on 2nd | R on 1st)
read one leaf
P(RR); P(RB)
add leaves, and maybe divide
P(R on 1st | B on 2nd); P(R on 2nd)
branch
A second-draw event given a first-draw condition is a forward conditional: 3 out of 10 after a blue, 8 out of 10 blue after a red. It is the label you wrote when you built the tree.
leaf
A named ordered pair is one complete path, so its probability is one leaf: 6 over 110 for two reds, 24 over 110 for red then blue.
work
Red second collects a leaf from each subtree, 30 over 110. Red first given blue second is backward: it collects the two blue-second leaves, 80 over 110, and divides 24 over 110 by that.

The dividing line is direction. A question that follows the tree, earlier stage to later stage, is on a branch or a leaf. A question about the later stage alone, or one that conditions on the later stage, cuts across the subtrees and needs a sum.

It is worth doing this sort in your head every time, because it tells you how much work to expect. If you find yourself dividing to answer a forward conditional, you have probably used the wrong tree; if you read a backward conditional off a branch, you have certainly used the wrong number.

41. Which supplier made the faulty part?

Real world

A factory buys 80 percent of a component from supplier A, whose parts are defective 1 percent of the time, and 20 percent from supplier B, whose parts are defective 5 percent of the time. A part fails inspection. Which supplier probably made it?

Figure (svg): A probability tree for parts from two suppliers. First branches A: 0.80 and B: 0.20. After A: defective 0.01 and fine 0.99. After B: defective 0.05 and fine 0.95. Leaves: A AND D 0.008, A AND fine 0.792, B AND D 0.010, B AND fine 0.190. The two defective leaves are highlighted.

The small supplier's defective leaf is the bigger of the two, even though its first branch is only a fifth of the tree.

Discussion prompt

The tree runs from supplier to defect. The question runs backwards: given defective, which supplier? Collect the defective leaves, then divide. Say in one sentence what the answer means for the factory.

Your instinct probably said supplier A, because A makes most of the parts. The tree shows why that instinct is incomplete: the first branch favours A four to one, but the second branch favours B five to one, and the leaf is the product of both.

This backward reading of a tree is the heart of medical screening, spam filters and quality control, and later courses name it Bayes' theorem. Here it is nothing but the tree plus the conditional formula: collect the leaves that match the evidence, then ask what share of them came from each branch.

OpenStax Introductory Business Statistics 2e, §3.4 Contingency Tables and Probability Trees §3.4 — probability trees in a business setting

42. Venn diagrams

Section

Part 4

43. A Venn diagram: a box for everything, a circle for each event

Concept

A tree is for things that happen in order. A Venn diagram is for two properties that one outcome may or may not have. It consists of a box that represents the sample space, S, with a circle or oval for each event inside it.

Venn diagram — A picture of the outcomes of an experiment: a rectangle standing for the sample space S, and circles or ovals standing for events. An outcome in an event is drawn inside that event's circle.

Figure (svg): A rectangle labelled S holding two overlapping circles. Circle A contains 1, 2, 3, 4, 5 in its left part and shares 6 with circle B. Circle B's own part holds 7, 8, 9. Outside both circles, at the right of the rectangle, are 10, 11 and 12.

Each of the twelve outcomes sits in exactly one of four regions. The 6 sits in the lens because it belongs to both events.

The book's first example has twelve equally likely outcomes, 1 to 12, with A the numbers 1 to 6 and B the numbers 6 to 9. Each outcome is written in the one region where it belongs. The 6 is in both events, so it goes in the lens where the circles overlap.

Two circles cut the box into four regions: A only, both, B only, and neither. Every outcome lands in exactly one of them, which is what makes the diagram useful. Any event built from A and B, their AND, their OR, a complement, is a collection of whole regions.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 193 — Venn Diagram, Example 3.27, Figure 3.6

44. Example 3.27: reading AND, OR and neither from the diagram

Worked example

Outcomes 1 to 12 are equally likely. A is 1 to 6 and B is 6, 7, 8, 9. Find the chance of A AND B, of A OR B, and of neither.

\[ A = \{1, 2, 3, 4, 5, 6\}, \qquad B = \{6, 7, 8, 9\} \]

Each answer is a set of regions, and since the twelve outcomes are equally likely, each probability is the number of outcomes in those regions over 12.

A AND B: the lens

Why: only 6 is in both circles.

\[ A\text{ AND }B = \{6\}, \qquad P(A\text{ AND }B) = \frac{1}{12} \]

A OR B: everything inside a circle

Why: the three regions A only, lens and B only.

\[ A\text{ OR }B = \{1, 2, \dots, 9\}, \qquad P(A\text{ OR }B) = \frac{9}{12} = 0.75 \]

Neither: the region outside both circles

Why: 10, 11 and 12.

\[ P(\text{neither}) = \frac{3}{12} = 0.25 \]

The diagram also shows why the OR is not simply 6 plus 4 outcomes. The 6 is written once, in the lens, so when you count the numbers inside the circles you count it once. The picture enforces the subtraction that the formula asks you to remember.

Notice that neither is the complement of the OR: the region outside the circles is everything not inside one of them. That is why 0.75 and 0.25 add to 1.

Check the OR with the addition rule

Why: A has 6 outcomes and B has 4, but the 6 is in both, so it must be taken away once.

\[ \frac{6}{12} + \frac{4}{12} - \frac{1}{12} = \frac{9}{12} \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 193 — Example 3.27

45. Example 3.28: two coins in a Venn diagram

Worked example

Flip two fair coins. A is tails on the first coin, B is tails on the second. Place the four outcomes and find the chance of A OR B.

List each event

Why: a pair belongs to A when its first letter is T, to B when its second is.

\[ A = \{TT, TH\}, \qquad B = \{TT, HT\} \]

The overlap

Why: TT is the only pair with tails on both coins.

\[ A\text{ AND }B = \{TT\} \]

Figure (svg): A rectangle for the four outcomes of two coin flips, with overlapping circles A, tails on the first coin, and B, tails on the second. TH sits in A only, TT in the overlap, HT in B only, and HH outside both circles in the corner.

Four outcomes, four regions, one outcome each. HH is in the rectangle but in neither circle: it is the complement of A OR B.

One outcome per region: TH in A only, TT in the lens, HT in B only, and HH outside both circles. The book calls the sample space X in its text and S in its figure; they are the same four outcomes.

A OR B: the three outcomes inside a circle

Why: four equally likely outcomes, three of them in the union.

\[ P(A\text{ OR }B) = \frac{3}{4} = 0.75 \]

In context: with two fair coins, at least one lands tails three times in four.

Check with the neither region

Why: HH is the only outcome in neither A nor B, and one minus its probability must match.

\[ 1 - P(HH) = 1 - \frac{1}{4} = \frac{3}{4} \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 194 — Example 3.28, Figure 3.7

46. Try It 3.28: place the die faces

Matching

Roll a fair die. A is a prime number of dots, 2, 3 or 5; B is an odd number of dots, 1, 3 or 5. Match each region of the Venn diagram with the faces that belong in it.

Match the pairs

  • a. inside A only
  • b. the overlap, A AND B
  • c. inside B only
  • d. outside both circles
  • w. 2
  • x. 3 and 5
  • y. 1
  • z. 4 and 6

Why: Three and 5 are both prime and odd, so they sit in the overlap. Two is prime but even, so it is in A only. One is odd but not prime, so it is in B only. Four and 6 are neither prime nor odd, so they sit outside both circles.

Fill the overlap first. Ask of each face, is it in both events? Only then place the rest. If you start with A only, it is tempting to write all of A, 2, 3 and 5, in the crescent, and then 3 and 5 are drawn twice.

\[ P(A\text{ OR }B) = \frac{4}{6}, \qquad P(A\text{ AND }B) = \frac{2}{6} \]

With the regions filled, the book's answers fall out: A AND B is 3 and 5, and A OR B is 1, 2, 3 and 5. Try It 3.27 works the same way with colours: blue is the only colour in both C and P, so blue goes in the lens and black, white and orange go outside.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 194 — Try Its 3.27 and 3.28

47. Reading the four regions piece by piece

Notation

When a Venn diagram carries probabilities instead of outcomes, each region holds one number. These three relations are all you need to fill and read it.

Annotate

On: \( P(A\text{ OR }B) = P(A \text{ only}) + P(A\text{ AND }B) + P(B \text{ only}) \)

  • The crescent of A outside B. It equals the probability of A minus the overlap, because the whole circle A is its crescent plus the lens.
  • The lens, counted once. Fill it first: every other region is found by subtracting it.
  • The crescent of B outside A: the probability of B minus the overlap.
  • Three regions, no double counting. Substitute the two crescents and this becomes the addition rule of Section 3.3.

Substitute the crescents and watch the addition rule appear. The lens is added once in the sum and subtracted twice inside the crescents, so on balance it is subtracted once.

\[ (P(A) - P(A\text{ AND }B)) + P(A\text{ AND }B) + (P(B) - P(A\text{ AND }B)) \]

\[ = P(A) + P(B) - P(A\text{ AND }B) \]

This is why the order of filling matters so much. The given numbers are usually the two circle totals and the lens. Only the lens is a region; the circle totals are sums of two regions each, so they must be broken apart before anything is written in a crescent.

The fourth region, outside both circles, is one minus that union. All four regions add to 1, and that sum is the Venn diagram's built-in check, just as the leaves were the tree's.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 195 — Example 3.29: P(C OR PT) = P(C) + P(PT) − P(C AND PT)

48. Example 3.29: clubs and part-time work

Worked example

Forty percent of the students at a college belong to a club, 50 percent work part time, and 5 percent do both. C is belongs to a club, PT is works part time. Fill the Venn diagram and find the chance of a club given part time, and of a club or part time.

\[ P(C) = 0.40, \quad P(PT) = 0.50, \quad P(C\text{ AND }PT) = 0.05 \]

Fill the lens first

Why: the both figure goes in the overlap as given.

\[ \text{lens} = 0.05 \]

Club only

Why: the club circle minus the lens.

\[ 0.40 - 0.05 = 0.35 \]

Part time only

Why: the part-time circle minus the lens.

\[ 0.50 - 0.05 = 0.45 \]

Neither

Why: whatever is left of the box.

\[ 1 - (0.35 + 0.05 + 0.45) = 0.15 \]

Figure (svg): A rectangle with two overlapping circles: C, club, 0.40, and PT, part time, 0.50. The overlap holds 0.05, the club-only part 0.35, the part-time-only part 0.45, and the space outside both circles 0.15.

The circle totals 0.40 and 0.50 are not region values. Each is the overlap plus a crescent, so the crescents are what is left after the 0.05.

Club given part time

Why: restrict to the part-time circle and take the lens's share of it.

\[ P(C \mid PT) = \frac{0.05}{0.50} = 0.1 \]

Club OR part time

Why: the three regions inside the circles.

\[ P(C\text{ OR }PT) = 0.35 + 0.05 + 0.45 = 0.85 \]

Notice how the conditional was found. Given part time restricts you to the part-time circle, 0.50 in all, and the students in it who also belong to a club are the lens. It is the same fraction-of-a-piece reading as on the tree: the AND over the condition.

In context: 85 percent of students belong to a club, work part time, or both, and only one part-timer in ten belongs to a club.

Check with the book's addition rule

Why: the two routes must agree.

\[ 0.40 + 0.50 - 0.05 = 0.85 \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, pp. 194-195 — Example 3.29, Figure 3.8

49. Example 3.29 drawn to scale

Picture it

The book's circles are not drawn to scale; they only show which region is which. Here the same four regions are laid end to end along a line of length 1, so each one's length is its probability.

Figure (svg): One horizontal bar of length 1, split into four pieces drawn to scale: club only 0.35, both 0.05 (yellow), part time only 0.45, neither 0.15. A bracket above spans club, 0.40; a bracket below spans part time, 0.50, and the yellow piece is one tenth of it.

Squeezed onto one line, the four regions fill exactly 1. The conditional probability of a club given part time is the yellow piece measured against the PT bracket: 0.05 out of 0.50.

Now the conditional is something you can see. Given part time means you stand inside the lower bracket, of length 0.50. The yellow piece, the students who also belong to a club, is 0.05 of that length: one tenth.

\[ P(C \mid PT) = \frac{0.05}{0.50} = 0.1, \qquad P(C) = 0.40 \]

Compare the two numbers. Across all students, 40 percent belong to a club; among part-timers, only 10 percent do. Working part time and joining a club are strongly dependent here, which makes sense: students with jobs have less spare time.

The product test agrees. If the events were independent, both would have probability 0.40 times 0.50, which is 0.20. The actual overlap, 0.05, is a quarter of that.

\[ P(C)\,P(PT) = 0.20 \ne 0.05 = P(C\text{ AND }PT) \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 195 — Example 3.29

50. Try It 3.29: second jobs and working spouses

Faded example

At a factory, 50 percent of workers have a second job, W, 25 percent have a spouse who also works, S, and 5 percent have both. Fill in the regions and the union, following the steps of Example 3.29.

Fill in the blanks

\text0.45 = 0.50 - 0.05 = 0.20, \quad \text0.70 = 0.25 - 0.05 = ___, \quad P(W\text___S) = ___

Why: Each crescent is its circle minus the lens: 0.45 for a second job only and 0.20 for a working spouse only. The union is the two crescents plus the lens, 0.45 plus 0.20 plus 0.05, which is 0.70.

The order is the one from the previous slides: lens, then crescents, then union, then the outside. The outside region here is one minus 0.70, so 30 percent of the workers have neither a second job nor a working spouse.

\[ 0.45 + 0.05 + 0.20 + 0.30 = 1 \]

Always finish with that sum. If the four regions do not total 1, a crescent was filled with a circle total, and the rest of the diagram is wrong with it.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 195 — Try It 3.29

51. Example 3.30: type O blood and the Rh-negative factor

Worked example

Four percent of African Americans have type O blood and a negative Rh factor, 5 to 10 percent have the Rh-negative factor, and 51 percent have type O blood. O is type O, R is Rh-negative. Find the four probabilities and describe two regions in words.

One number is a range, not a value. The book takes the average of 5 and 10 percent, and so will we; say so whenever you report an answer built on it.

(a), (b), (c): the given numbers

Why: averaging the range for Rh-negative.

\[ P(O) = 0.51, \quad P(R) = \frac{0.05 + 0.10}{2} = 0.075, \quad P(O\text{ AND }R) = 0.04 \]

(d) The addition rule

Why: the events overlap, so subtract the lens once.

\[ P(O\text{ OR }R) = 0.51 + 0.075 - 0.04 \]

Evaluate

Why: matches the book's 0.545.

\[ P(O\text{ OR }R) = 0.545 \]

Figure (svg): A rectangle with a large circle O, type O blood, 0.51, and a smaller oval R, Rh-negative, 0.075, overlapping at the right of the circle. The overlap holds 0.04, the O-only part 0.47, the R-only part 0.035, and the outside 0.455.

The oval is small, and more than half of it lies inside the circle: most Rh-negative people in this population also have type O blood.

(e) The overlap represents the African Americans who have type O blood and the Rh-negative factor: the universal donors. (f) The region in the rectangle but outside both shapes represents those who have neither type O blood nor the Rh-negative factor.

Fill the other regions

Why: each crescent is its shape minus the lens; the outside is one minus the union.

\[ 0.51 - 0.04 = 0.47, \quad 0.075 - 0.04 = 0.035, \quad 1 - 0.545 = 0.455 \]

Check that the four regions total 1

Why: lens, two crescents and outside.

\[ 0.47 + 0.04 + 0.035 + 0.455 = 1 \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, pp. 195-196 — Example 3.30, Figure 3.9

52. Trap: writing the whole circle in the crescent

Trap

The trap

Filling Example 3.29, a common first diagram puts the circle totals straight into the crescents:

\[ \text{C only} = 0.40 \]

\[ \text{PT only} = 0.50 \]

\[ \text{lens} = 0.05 \]

Wrong. The three regions add to 0.95, leaving 0.05 for neither, when the true answer is 0.15.

The fix

A circle's total includes its lens. So the crescent is a subtraction, done after the lens is filled:

\[ 0.40 - 0.05 = 0.35 \]

\[ 0.50 - 0.05 = 0.45 \]

The first diagram gave itself away: its regions did not total 1 once the outside was checked against the data. Always compute the outside last, as one minus the three inner regions, and ask whether it is believable.

Then the regions total 0.85 and neither is 0.15. The habit that prevents this: lens first, crescents second, and the four-region sum as the last line.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 195 — Example 3.29

53. Try It 3.30: novels, non-fiction, and a CD oval

Worked example

In a bookstore, a customer buys a novel with probability 0.6, a non-fiction book with probability 0.4, and both with probability 0.2. Draw the diagram, find the chance of either kind of book, and add an oval for customers who buy only CDs.

Lens first

Why: both kinds of book.

\[ P(N\text{ AND }NF) = 0.2 \]

Crescents

Why: each circle minus the lens.

\[ \text{novel only} = 0.6 - 0.2 = 0.4, \quad \text{non-fiction only} = 0.4 - 0.2 = 0.2 \]

(b) Either kind of book

Why: the addition rule.

\[ P(N\text{ OR }NF) = 0.6 + 0.4 - 0.2 = 0.8 \]

Notice that the two circle totals, 0.6 and 0.4, add to exactly 1, yet 20 percent of customers are outside both circles. That is only possible because the circles overlap by 0.2: adding the circles counted those customers twice.

Figure (svg): A rectangle with overlapping circles N, novel, 0.6, and NF, non-fiction, 0.4. The overlap holds 0.2, the novel-only part 0.4 and the non-fiction-only part 0.2. In the lower right, a small separate oval labelled CD only touches neither circle. The outside region is labelled 0.2.

The CD-only oval sits in the outside region and touches neither circle: those customers bought no book of either kind.

(c) The overlapping area represents the customers who buy both a novel and a non-fiction book. (d) Customers who buy only CDs buy no book at all, so their oval sits in the outside region and touches neither circle. Its size is not given; it is some part of the outside 0.2.

Check that the regions total 1

Why: two crescents, the lens, and the outside.

\[ 0.4 + 0.2 + 0.2 + 0.2 = 1 \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 196 — Try It 3.30

54. Can any three numbers make a Venn diagram?

Counterexample

Someone hands you P(A) of 0.5, P(B) of 0.6, and P(A AND B) of 0.05, and asks you to draw the Venn diagram. Try it before reading on.

Discussion prompt

Fill the four regions with these numbers and show that they cannot all be right. What is the smallest the overlap could be?

A Venn diagram is a check on data as well as a way to read it. Two circles whose totals add to more than 1 are forced to overlap by at least the excess, because the box has only 1 to give. Here the excess is 0.1, and 0.05 is too small.

\[ P(A\text{ AND }B) \ge P(A) + P(B) - 1 = 0.1 \]

The bookstore sat exactly on the edge of this: its circle totals added to 1, so the overlap had to be at least zero, and any overlap, like its 0.2, left room outside. Whenever two totals add to more than 1, run this check before drawing anything.

55. Neither and not both are different regions

Intuition

Two complements come up constantly with Venn diagrams, and their names sound alike. The picture keeps them apart.

Neither A nor B is the region outside both circles: the outcomes in no event at all. It is the complement of the union, the OR.

\[ P(\text{neither}) = 1 - P(A\text{ OR }B) \]

Not both is everything except the lens: the outside region and both crescents. It is the complement of the intersection, the AND, and it includes the outcomes where exactly one event happens.

\[ P(\text{not both}) = 1 - P(A\text{ AND }B) \]

For Example 3.29 the two are very different: neither is 1 minus 0.85, so 0.15, while not both is 1 minus 0.05, so 0.95. A student who is in a club but has no job is in the not-both region and not in the neither region.

When a question uses one of these phrases, shade the region on the diagram before computing. Shading makes you decide what the words mean, and the region you shaded tells you which complement to take.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 193 — Example 3.27: A OR B and the region outside it

56. Exercise 120: filling the diagram from a union

Worked example

Given events J and K with P(J) of 0.18, P(K) of 0.37 and P(J OR K) of 0.45, find P(J AND K), the probability of the complement of J AND K, and the probability of the complement of J OR K.

This time the union is given and the lens is not. The addition rule has four quantities in it, and any three give you the fourth, so solve it for the lens.

Write the addition rule

Why: the general form, always true.

\[ P(J\text{ OR }K) = P(J) + P(K) - P(J\text{ AND }K) \]

Solve for the lens

Why: move the AND to one side and the union to the other.

\[ P(J\text{ AND }K) = P(J) + P(K) - P(J\text{ OR }K) \]

(a) Substitute

Why: the two circle totals and the union.

\[ P(J\text{ AND }K) = 0.18 + 0.37 - 0.45 = 0.10 \]

(b) Complement of the lens

Why: everything except the overlap.

\[ 1 - 0.10 = 0.90 \]

(c) Complement of the union

Why: the region outside both circles.

\[ 1 - 0.45 = 0.55 \]

Now every region is known: J only is 0.08, the lens 0.10, K only 0.27, and the outside 0.55. The two complements in parts (b) and (c) are very different, because not both and neither are very different events.

Check that the four regions total 1

Why: crescents, lens, and outside.

\[ 0.08 + 0.10 + 0.27 + 0.55 = 1 \]

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3 Bringing It Together homework, p. 216 — Exercise 120

57. Choosing a display

Section

Part 5

58. Sequence, overlap, or a grid of counts

Concept

You now have three displays for probability: the contingency table of Section 3.4, the tree, and the Venn diagram. They are not rivals. Each one makes a different kind of question easy, and the choice is decided by the shape of the problem.

displayuse it whenwhat it makes free
tree diagramthings happen in stages: draw, then draw again; supplier, then defectthe multiplication rule, forward conditionals, with against without replacement
Venn diagramone outcome may have two properties at onceunions, overlaps, complements, the addition rule
contingency tablecounts of two categorical variables for a group of people or itemsevery marginal and conditional as a row or column share

Listen for time in the wording. First, then, second, after, one at a time: that is a tree. Belongs to, has, both, either: that is a Venn diagram. A table of counts classified two ways is already a contingency table, and turning it into anything else loses nothing but also gains little.

The deeper point is that the three are the same information arranged differently. A two-stage tree has four leaves, a two-event Venn diagram has four regions, and a two-by-two table has four cells. Next slide: one problem, all three ways.

OpenStax Introductory Business Statistics 2e, §3.4 Contingency Tables and Probability Trees §3.4 — contingency tables and probability trees side by side

59. One problem, three pictures: Example 3.29 as a tree

Concept

Take the clubs and part-time work from Example 3.29 and redraw them as a tree, club membership first. The first branches are the given 0.40 and its complement 0.60. The second branches need conditionals, which the Venn diagram only held implicitly.

Part time given club

Why: the lens as a share of the club circle.

\[ P(PT \mid C) = \frac{0.05}{0.40} = 0.125 \]

Part time given no club

Why: the part-time crescent as a share of everything outside the club circle.

\[ P(PT \mid C') = \frac{0.45}{0.60} = 0.75 \]

Figure (svg): A probability tree for Example 3.29. First branches C: 0.40 and not C: 0.60. After C: PT 0.125 and not PT 0.875. After not C: PT 0.75 and not PT 0.25. Leaves: C AND PT 0.05, C AND not PT 0.35, not C AND PT 0.45, neither 0.15.

The four leaves are the four Venn regions. The second-stage labels are new information the Venn diagram only held implicitly: 0.05 divided by 0.40 and 0.45 divided by 0.60.

The four leaves reproduce the four Venn regions exactly: 0.40 times 0.125 is the lens, 0.60 times 0.75 is the part-time crescent. Nothing is lost in translation, but the tree shows something new at a glance: three quarters of non-members work part time, against one in eight members.

PTnot PTtotal
C0.050.350.40
not C0.450.150.60
total0.500.501

And the same four numbers make a two-by-two table. Its cells are the regions, its margins are the circle totals. Choose the picture that puts the number you need where you can read it.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3.5, p. 195 — Example 3.29, redrawn

60. Which display fits the problem?

Sorting

For each problem, choose the display you would draw first. More than one could work; pick the one that makes the question easy.

Sort into buckets

Sort each problem by the display that suits it best.

tree diagram
two cookies eaten one after the other from a box; a part from one of two suppliers, then found defective or not
Venn diagram
the share of students who are in a club, work part time, or both; customers who buy a novel, a non-fiction book, or both
contingency table
counts of smokers by ethnicity and cigarettes per day; 400 patients classified by treatment group and by outcome
tree
Cookies and parts both happen in stages: the second cookie depends on the first, and the defect depends on the supplier. A tree puts each stage's conditional on its own branch.
venn
Clubs and work, novels and non-fiction are two properties one person can have at once, with no order. The overlap is the whole question, so draw it.
table
Both come as counts cross-classified by two categorical variables. The counts already fill a grid, and every probability is a cell over a total.

The giveaway for a tree is dependence in time: something happens, and then something else happens with chances that may depend on it. The giveaway for a Venn diagram is the word both, with no order attached. Counts in two directions are a table.

Notice that the smoking counts come from the book's practice exercises for this chapter, and the book puts them straight into a table. You could draw a tree with five ethnicity branches and four smoking branches on each, but twenty leaves is exactly the kind of clutter a table avoids.

61. Put the tree routine in order

Ranking

Here is the routine the tree examples followed. Put the steps in the order you would carry them out on a new two-stage problem.

Put in order

Order the steps for answering a question with a probability tree.

  1. Decide the stages and whether the draw is with or without replacement
  2. Label the first-stage branches with plain probabilities
  3. Label each second-stage branch by asking what is left now
  4. Multiply along every path and check that the leaves total 1
  5. Choose the leaves the event uses, and add them
  6. For a conditional, divide by the leaves of the condition, then state the answer in context

Why: The replacement question comes first because it decides every second-stage label. The leaves are checked before they are used, so an error in a branch is caught before it spreads into every answer. Conditionals come last because they are built from sums of leaves.

The step people skip is the fourth. Checking that the leaves total 1 takes one line and catches most slips in the branch labels, such as a pair of second-stage branches that does not add to 1. It cannot catch the with-replacement trap, since that tree also totals 1; only the first step, deciding about replacement, protects you there.

The step people rush is the fifth. Name every leaf the event uses, in both orders, before adding. One of each colour is two leaves; red second is a leaf from each subtree.

62. Putting it together

Section

Part 6

63. Pattern: draw it, check it, read it

Pattern

Both displays follow the same three moves. Draw the structure from the problem, check it against a total, then read every answer as a collection of pieces.

tree diagramVenn diagram
drawa column of branches per stage; second-stage labels are conditionalsa box for S, a circle per event; fill the lens first
checkleaves total 1, or the size of the sample spacefour regions total 1
ANDone leaf, the product along the paththe lens
ORadd the leaves the event usesadd the regions inside the circles
conditionalforward: a branch label; backward: leaves over leavesthe lens as a share of the condition's circle
complementone minus the leaves left outthe regions outside the event
  1. Ask what is left before every second-stage label; without replacement, the denominator drops by one.
  2. Fill the overlap first; every crescent is a circle total minus the lens.
  3. Name every leaf or region an event uses before adding.
  4. Check the total before answering anything.

Read the table by rows as much as by columns. An AND is always a single piece, a leaf or the lens. An OR is always a sum of pieces. A conditional is always a ratio of pieces, and a complement is always what is left of the whole. Once you can point at the pieces, the formulas become bookkeeping.

The rules of Section 3.3 have not gone away. They are what the pictures compute: a path is the multiplication rule, the three inner regions are the addition rule, and a conditional is still the AND over the condition.

64. Check: a tree without replacement

Check

One path, one product. Decide what is left in the deck before writing the second factor.

Check your understanding

You have 8 cards, 5 green and 3 yellow. You draw two without replacement. What is the probability that both are green?

  • A. 20/56 (correct)
  • B. 25/64
  • C. 25/56
  • D. 10/56

Answer: A

Why: The first branch is 5 green out of 8. Given a green is gone, 4 green remain among 7, so the second branch is 4 out of 7. The path probability is 5 times 4 over 8 times 7, which is 20 over 56, about 0.357.

Why B tempts people
That is 5 over 8 squared, the with-replacement answer. The first card is not put back, so the second draw is from 7 cards with 4 green.
Why C tempts people
That uses the right denominator of 7 but keeps 5 green, as if the green card drawn first were still in the deck.
Why D tempts people
That is half the right answer, as if the green-green path had to be split between two orders. A single path has only one order.

The quickest route is to draw only the path you need. The first branch is 5 green out of 8 cards; the second is out of 7, with one green gone. There is no need for the other three leaves unless you want the total check.

Say the second branch in full: green on the second draw, given green on the first. The condition tells you a green card is gone, which changes both the numerator and the denominator.

This is Exercise 122(b) from the chapter homework. With replacement, Exercise 121 gets 25 over 64, about 0.391, a little higher: removing a green makes the second green less likely.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3 Bringing It Together homework, p. 216 — Exercises 121-122

65. Check: filling a Venn diagram

Check

Lens first, then crescents. Decide what goes in the region before choosing.

Check your understanding

P(A) = 0.5, P(B) = 0.3 and P(A AND B) = 0.1. What probability belongs in the region inside A but outside B?

  • A. 0.4 (correct)
  • B. 0.5
  • C. 0.2
  • D. 0.7

Answer: A

Why: The circle A holds 0.5 in all, and 0.1 of that is the lens it shares with B. The crescent is what is left: 0.5 minus 0.1, which is 0.4.

Why B tempts people
That is the whole circle A, lens included. Writing it in the crescent counts the lens twice and makes the regions total more than they should.
Why C tempts people
That is the B-only crescent, 0.3 minus 0.1. The question asks for the part of A outside B.
Why D tempts people
That is the union, 0.5 plus 0.3 minus 0.1: all three regions inside the circles, not one crescent.

The circle A is made of two regions, the crescent and the lens, and the given 0.5 is their sum. Take the lens away and the crescent is what remains; that is the whole calculation, and it is the same move you made for the clubs and for the blood types.

Once the A-only crescent is 0.4 and the B-only crescent is 0.2, the outside region is one minus 0.7, which is 0.3. Checking: 0.4, 0.1, 0.2 and 0.3 total 1.

If you chose the whole circle, go back to the trap in Part 4. A circle total is never a region value unless the circles do not overlap.

66. Check: a backward conditional

Check

Use Miguel's cookie tree: 3 chocolate and 7 butter, two eaten without replacement. The leaves are 6, 21, 21 and 42, all over 90.

Check your understanding

Given that Miguel's second cookie was butter, what is the probability that his first cookie was chocolate?

  • A. 1/3 (correct)
  • B. 7/9
  • C. 21/90
  • D. 3/10

Answer: A

Why: The condition, butter second, collects the chocolate-butter leaf, 21 over 90, and the butter-butter leaf, 42 over 90, for 63 over 90. The chocolate-first part of that is 21 over 90. Dividing gives 21 over 63, which is one third.

Why B tempts people
That is the branch label for butter second given chocolate first. It is the forward conditional; the question conditions on the second cookie, so it runs backwards.
Why C tempts people
That is the AND, chocolate first and butter second, without dividing by the probability of the condition.
Why D tempts people
That is the plain chance of chocolate first. Learning the second cookie was butter changes it slightly, from 0.3 to about 0.333.

The condition is about the later stage, so it takes a leaf from each subtree. That is the signal that the answer is not on any branch and a division is coming.

Notice the answer, one third, is 3 out of 9: given the second cookie was butter, the first is like a draw from the other 9 cookies, 3 of them chocolate. Order does not matter to the chances, only to the arithmetic.

67. Explain why the second branch is a conditional

Explain it to yourself

The single idea that makes trees work is that the second-stage labels are conditional probabilities. Put that in your own words.

Discussion prompt

In two or three sentences, explain why every second-stage branch on a probability tree is a conditional probability, and why that makes the product along a path equal to the probability of the AND.

A good explanation names the position of the branch. It is drawn after the first branch, so it can only be read after the first event is known; that is what a conditional means.

Test your explanation on the replacement question. If it explains why the second labels change without replacement and stay the same with it, it has understood what the labels are.

A strong answer also says what this buys you: you never have to remember to use a conditional, because you cannot label a second-stage branch without asking what already happened.

68. Exit ticket

Exit ticket

You have 8 cards, 5 green and 3 yellow, and draw two without replacement. G1 is a green first card, G2 a green second card.

Discussion prompt

Build the tree. Then find the probability of at least one green, the probability of G2 given G1, and the plain probability of G2. Say whether G1 and G2 are independent, with the numbers that decide it.

Every part of the lesson is in this one problem: a tree without replacement, a complement read as one leaf, a forward conditional read off a branch, and a plain second-draw probability collected from both subtrees.

Notice that the plain chance of green second is 5 out of 8, the same as the first draw, just as the urn and the cookies showed. Dependence lives in the conditional, not in the plain probability.

OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams §3 Bringing It Together homework, p. 216 — Exercise 122

69. Recap

Recap

Two pictures, each drawing one of the rules of Section 3.3 so that the rule is hard to misuse.

tree diagramVenn diagram
drawsan experiment in stagestwo events that can overlap
built-in rulepath = product: P(A) · P(B | A)union = A only + lens + B only
watch forsecond stage changes without replacementcrescent = circle − lens
checkleaves total 1regions total 1

\[ P(A \mid B) = \frac{P(A\text{ AND }B)}{P(B)} \quad \text{on either picture} \]

Forward conditionals are free on a tree; backward ones collect leaves and divide. On a Venn diagram, a conditional is the lens as a share of the condition's circle. And the tree, the Venn diagram and the contingency table hold the same four numbers for two events: choose the one that puts your answer in plain sight.

OpenStax Introductory Business Statistics 2e, §3.5 Venn Diagrams §3.5 — Venn diagrams with business examples

Sources

  1. OpenStax Introductory Statistics 2e, §3.5 Tree and Venn Diagrams — Illowsky & Dean, OpenStax / Rice University, CC BY 4.0, pp. 188-196
  2. OpenStax Introductory Business Statistics 2e, §3.5 Venn Diagrams — Illowsky & Dean, OpenStax / Rice University, CC BY 4.0
  3. OpenStax Introductory Business Statistics 2e, §3.4 Contingency Tables and Probability Trees — Illowsky & Dean, OpenStax / Rice University, CC BY 4.0

Want this taught 1-on-1? Alexander tutors Statistics — $55/session, free consultation.

Book on Wyzant · Text (657) 465-8108