AP Statistics

Chapters 10 & 11 Review solutions

Chapter 10:

1 C (Use pat .5 because you aren't told what to use, and .5 maximizes the numerator of the standard

deviation. 1. 96 [ ~p(1: p) }; 0.08 So n e 150.0625 ..... therefore choose the closest sample size above 150.) 2. Increase the width of a confidence interval by : increasing sample size, decreasing standard deviation, or decreasing confidence.

3 To cut the margin of error in half, quadruple the sample size. M = z * (_!!_J so _!_ M = _!_ z * (_!!_J or,

.i; 2 2 ..In

moving the 1 to inside the square root: l M = z * ( {;- J

2 2 ,,4n

4 Use a t-CI instead of a z-CI whenever you don't know the population standard deviation

5 We are using a method that constructs an interval that contains the true mean (or proportion) 95% ofthetime. 6 Either a significance test or a confidence interval would work for this. See the attached scoring guidelines fromAP.

Chapter 11: lA

2 B (you can either calculate this by nonnalcdf(1.40, some big munber, 0, 1)01' look up the z-score of 1.40 in the table and find (J-that area)) Because we don't have an actual standard deviation or data to use, you cannot calculate this based on a sample. You must use the z-score (thusly, in your normalcdf the mean of all z-scores is 0, and one standard deviation using z-scores is 1)

3 D (you can either calculate the normalcdf(0.8, some big munber, 0, 1) and double it to account for the twosided p-value or you can look up 0.8 in your normal table and find (1- that area) and double it.)

4 D because p = 0.08 > 0.05 = a

5 E because you don't know EXACTL Y what the p-value was for the test. Knowing that the null hypothesis can be rejected at the 5% level of significance only means that you know the p-value is somewhere between 0 and 5%. It might be less than 1 % (at which point you could reject it at the 1 % level) or it might not be. We do not have enough information to answer this. B is wrong because it implies that we know the p-value and it is below 5%, but above 1 %, therefore, not strong enough "evidence" to reject.the null hypothesis .... but we don't know this.

6B

7 A Start by writing what a Type II error is: You did not reject the null hypothesis, but should have. Then, translate that into the context of the problem: The examiner said the person was innocent, but they were actually guilty. There were 15 cases that fell into that category. We are asking the Type II error, which is the area under the graph of the ENTIRE true population, so take the 15 out of the entire population.

8 A p-value is the area under the curve from the sample mean (or proportion) value to the tail (or that area doubled, if two-sided). More specifically (and what AP would want): The probability of getting a sample as far or further (in one or both directions) from the expected parameter in a population that actually looked like the expected parameter.

9 P = 1- j3 It is the strength (to eliminate type II error) of a significance test against a proposed alternative parameter value. Increase power by: decreasing type II error, increasing type I error, increasing sample size, decreasing standard deviation, or by choosing the alternate parameter value further from the expected parameter value.

10 Type I Error: you reject the null hypothesis, but you shouldn't have P(Type 1 error) = a

Type II Error: you fail to reject the null hypothesis, but you should have P(Type 11 error) = j3

11 Type I: Reject the day's production, which results in a loss of money for the manufacturer, and a waster of perfectly fine supplies because the bulbs were actually fine.

Type II: Ship the day's bad bulbs out to stores, which results in consumers being unhappy with the product and therefore the manufacturer

12 Complete the steps for either a confidence interval or a significance test. (The book refers to these steps as the "Inference Tool box" and includes stating the procedure (and hypotheses if doing a test), checking conditions, doing the calculations, and interpreting the results as a conclusion.)

Conclusion: There is significant evidence to conclude that the mean time to complete the maze is changed by vigorous exercise because the p-value is 0, which is clearly less than level of significance (a =whatever you chose). I reject the null hypothesis that the mean time is still 30 seconds.

Conclusion: There is significant evidence to conclude that the mean time to complete the maze is changed by vigorous exercise because at the 95% (or whatever you choose) level of confidence the expected true mean for completion time lies between 0 and 22.72 seconds, which does not include the mean of30 seconds.

(Please note the interval is actually (-16.48,22.72) but negative times clearly don't make sense.)

13 Complete the steps for a significance test. (The book refers to these steps as the "Inference Tool box" and includes stating the procedure and hypotheses, checking conditions, doing the calculations, and interpreting the results as a conclusion.)

Conclusion: I reject the null hypothesis that the mean DRP score of all third graders in the district is at the nation mean of 32 at the 5% level of significance because the p-value for the signifance test based on our sample was 0.03, which is less than 0.05. There is significant evidence that the third-graders in this district have a mean score higher than the national mean.

14 See the attached scoring guidelines from AP.

15 See the attached scoring guidelines from AP.

Question 5 Scoring Guidelines

4 Complete Response

A. Selects an appropriate test of significance.

Observes that data are paired by hour of day and indicates a paired t-test is appropriate OR observes that the data are paired by hour of day and justifies choosing a two-sample t-test (or confidence interval for the difference of two proportions) OR worries about assumptions for a t-test and so chooses a

sign test OR performs a substantial data analysis.

B. States and addresses assumptions necessary for t-test, The dot plot below shows no evidence that the differences aren't normally distributed.

•

• • • •

e e

e

+---------+---------+---------+---------+---------+-------d

-5.0 -2.5 0.0 2.5 5.0 ",.7~5.

(Por a two-sample t-test, men~ions that outliers indicate a potential problem.)

C. Carries out a test of significance.

States the hypotheses correctly, with two-sided alternative, defining any nonstandard symbols.

Ho: !LA = !LB or!Ld = 0 H~ : !-LA =f= !-LB or !-Ld =f= 0

Calculates an appropriate test statistic, possibly

with trivial arithmetic error. .

d-O 1.11

t=---= 0.78

. . sdl-J;. 4.28/.J9.

(For two-samplet-test, t = 0.519 . and p = 0.61; pooled or unpooled. For a sign test, +-+-++-0+ so P = 0.727.)

Correctly approximates the P-value from a table or calculator.

With 8 degrees of freedom, the two-sided P-value is 0.46 (OR compares t = 0.'18 with critical value of 2.306).

OR constructs and interprets a 95% confidence interval for the mean difference:

OR a 95% confidence interval for the difference of two means: (-3.47, 5.69)

D. Correctly states a conclusion in the context of the situation.

e·

The difference in the mean percentages in the samples of defective chips produced by. ovens A and B isn't statistically significant. If there was no difference in the percentages, a mean absolute difference of 1.11 or greater would happen 46% of the time with random samples of size 9. "

Or, for the confidence interval: Since the confidence interval includes zero, there is no evidence that one oven differs from the other in terms of mean percentage of defective chips.

3 Substantial.Response

, .~ -_ ..

Perform~ a t~st of significance but

Doesn't mention the assumptions for the test used

OR

States an incorrect conclusion

OR

Makes two or three less significant omissions or errors inrhe test of significance, such as

• doesn't make it clear by words or formula which test they are using

• chooses 'atwo=sample t-test or. a sign-test-without

consideringthe paired t-test

• fails to list hypotheses

• doesa one-tailed test or forgets to double the P-value

• says must use a t-test because n is small

• uses z-test, not t-test

, • mentions normality assumption but doesn't check it graphically or appeal to robustness

• confuses parameters and statistics

• doesn't define nonstandard symbols

d ±t * ]; = 1.11 ±(2.306{ 798)= 1.11±3.29 or (-2. 18, 4.40) /r--7-.5--'· --,

Question 5 Scoring Guidelines (corrt.)

•

1

forgets {;; in formula for t

1 Minimal Response

Recognizes that the problem involves an inferential statistical argument.

Does little more than carry out the mechanics of a test . of significance, such as those done by a calculator.

•

uses wrong critical value or degrees of freedom "accepts" the null hypothesis or uses similar careless language in conclusion

gives correct, but minimal, conclusion ("do not reject H,") or a generic conclusion that doesn't relate to the ovens

Note: Since we are using holistic grading, a deficiency in one part can be compensated for by outstanding work in another.

•

2 Developing Response

Performs test of significance except makes both major errors OR an equivalent combination of errors.

Sample 1

S. A company bakes computer chips in two ovens, O'YCn A and oven B. The chips are randomly assigned to an oven and hundreds of chjps are baked each hour. The percentage of defective chips coming from these ovens for 'each hour' of production throughout a day is shown below,

Percentage of Defective Chips

Hour Oven A OvenB
1 45 36
2 32 37
3 34- 33
4 31 34
5 35 33
6 37 32
7 31 33
s 30 30
9 27 24 Theper~ntage of defective chips. produced each hour by oven A has a mean of 33.56 and a standard

. deviation of 5.20. The percentage of defective chips produced each hour by ovenB bas a mean of 32.44 and a standard deviation of'3.78.·The hourly differences in percentages for oven A minus oven B have a mean of 1.11 and a standard deviation of 4.28.

Does there appear to be a difference between oyen A and oven 13 with respect to the mean percentages of defective cbipsproduced? Give appropriate statis~ca1 evidence to support Jour IjDswer.

-r~ I~sl ~y" ~ ,··d,{{'e.rene.e h~f:",-·ee..,o._tP,..,4

alncf 6 ve~ 13 dtZ ~r;,Av.te !G-) l~~&, a n.;a.lt:-h ~ cI J;}et_·""..s

d €£ t!f} f1 vJ 1""/ b~ v:; ~ c/. jj, £. r e o s e t-r f~ :Q <.

'5, .( • .-, -/fer p /; .)' CO ;".. • ' .$ ~··o W,5 a 9 c c I .p C_:; .'/' ' .... ~

a 5 5oc../CI ,-<.~~ .~, b £/ IN- ~e.1<1' I- i; e p,erL.e"l./~ e

ot Ie fee...lrve ~ Q c.4 I~ s '. Tt. e_ prOt 60. t t 1'..;y

O ..... !# - A

76

'Sample 1 (cant.)

77

Sample 2

5. A company bakes computer. chips in two ovens, oven A and oven B. The chips are randomly assigned to an oven and hundreds of chips are baked each hour. The percentage of defective chips coming from

these ovens for each hour of production throughout a day is shown below. .

Percentage of Defective Chips

Hour Oven A OvenB
1 45 36
2 32 37
3 34 33
4 31 34
5 35 33
6 37 32 ..
7 31 33
8 30 30
9 27 24 The percentage of defective chips produced each hour by oven A has a mean of 33.56 and a standard deviation of 5.20. The percentage of defective chips produced each hour by oven B has a mean of 32.44 and a standard deviation- of 3.78. The hourly differences in percentages for oven A minus oven B

have a mean of Ll1 and a standard deviation of 4.28.· .

Does there appear to be a difference between oven A and oven B with respect to the mean percentages of defective chips produced? Give appropriate statistical evidence to support your answer .

'r-e d Il+ Q. \ 5 ~ ~ \It' 1\ -tC u.s 11\ MII:tc.. ~f~

P~lfs .forMI ""o.-t(.,,,,,~ d ~ ~I!l\)r\ l0 o:no..\'f-"2f

t-: 01t'n ~ 6 \' G.i r S d uta.1 1. \Nt \ \ «se C\ Ol\f --S:o.lV\ p~f

+ - t ~-\: (J(~N\ -t\\ 'f d \ f- forfl\( f '5 ':L \\ <:> .t~ s -to .

u. \.t? \>.l\' ~'f t"f )\o~ - \ r\d·t p6\d6\{t -rha+ MClt(h( d ~Q,.~rs ha..) '-

\-\:~ f..a,H = 0 ~~i{~~ ,,\\ ~-=- l\,:l<J f\ ~-C\

d.::.. lO~ .

d~::·t

--

78

-- '-.~-. .

c t,t(<l\ 1olf.C5 d~ 'J: ,30~ ..

J. \~'Ob '7 ,7 7~

B eCql\ tl ~e crif.t· (\1\ + (5 If'SS -HII'! ft-~ ;>~ .s~-tfs-t\()· "L t~m not (~j~{!;t He:> Iii fm,lJo'l of rtfl.' \.. ~ ,,~ 1\ 0 -e ~~ cl f'\ t~. i? ! '-I. S~ es.-t,~. ~ \'ft-er eN f i'V'

~~ pen.€h1Q. ~p or bo;,teo (~W; \·e~v l~ i\.,e oven,

. " ' . ,.: .

Sample 2 (cent.)

. :'l:

,I 79 I

Sample 3

S.A company bakes computer chips in two ovens. oven- A and oven B. The chips are randomly assigned . to an oven and hundreds of chips are baked each hour. The percentage of defective chips coming from these ovens for 'each hour of production throughout 'a -d~~_~_shown below.

Percentage of Defective Chips

Hour Oven A OvenB
1 45 36
2 32 37
3 34 33
4 31 ,., 34
5 35 33
6 - 37- - --32
.7 31 33
8 30 30
, 27 24 , The percentage of defective chips. produced each hour by oven A has a mean of 33.56 and a standard deviation of 5.20. The percentage of defective chips produced each hour by oven B bas. a mean of 32.44 and a standard deviation of 3.78. The hourly differences in percentages for oven A minus oven B

i have a mean of 1.11 and a standard deviation of 4.28. .

if

,

Does. there appear to be a difference between oven A and oven B with respect to the mean percentages of defective cbips produced? Give appropriate statistical evidence to support )'OUr answer.

See student response on facing page.

80

Sample 3 (cent.)

J s". 2<). 2. -J-~ 2-

. 'I . . g;

d:£ -:=. 1(. ~q2- ?- ~~ ~ 7i -9:2.-

po:; ,. Gt;9~}(S7" (.;r~ r)~~ ~/iA c-f? ---Iy_ ~e?f.f.e..J c:«:

en • "(~ ~ 9- t ~",.._ If?!-r--- -e..) , .

~ M 4dt~,A;y,Ae~p~'H,~

~ 'D ~ ~rfJlhl.~. ~.,A,D ~ ~ ~ f~+r:c.,,··· ~ 4-~ ~4

~ t!r~ A ~"' . .& '

~~.~-A~·~ ~iu ~ .

~ ..... seu~'- ,. •. ;p. M ~ ~ l' .c:t ~ ~ ~. --t-~"o~. l,.&)(....;~ ', ~ ~ ~ ~~ .qi'(..J.._.

Ap® STATISTICS

2003 SCORING GUIDELINES

Question 1

Solution

Part (a):

Min Q, Median Q3

Max

Students -4.5

~1.75 (or -0.5) o

1.0 (or 0.5) 5.0

no outliers

Teachers -2.0

-1.5

-1.0

o 0.5

no outliers

StudentS

Teachers

r-.-'-..----,--.--"l; _',..............

~ ~ ~ ~ ~ 0 1 2 3 4 5 Time Error

Part (b):

The teachers' watches tend to have tunes that are closer to the true tune. Although the teachers' watches tend to be slow, the tunes are less variable than the student times and so more teachers' watches had times that were close to the true tune.

Part (c):

. ····,'_·_-'Festing-this pair of hypotheses will not answer the teacher's question. The mean amount of tune by which student watches differ from the true tune could be zero even when student watch times differ greatly from the true time if large positive differences are offset by large negative differences.

Copyright © 2003 by College Entrance Examination Board. All rights reserved.

Available at apcentral.collegeboard.corn.

2

Ap® STATISTICS

2003 SCORING GUIDELINES

Question 1 (cont'd)

Scoring:

Each part is essentially correct (E), partially correct (P), or incorrect (I).

Part (a) is essentially correct (E) if parallel boxplots are correctly drawn (using either version of students' quartiles), includes correct labels (students vs. teachers) and shows a common scale. If only one group is labeled, we'll infer the correct label for the other group. A label on the scale is not needed, but should be correct if included. The five-number summary is not required, but may be used to compensate for an incorrectly placed point on a boxplot.

Part (a) is partially correct(p) if there is only one error (e.g., omitting labels and/or scale, showing outliers, highly inconsistent scale, one misplaced point, an incorrect method of calculating quartiles, or reversing labels). Minor arithmetic errors will be overlooked.

NOTE: Two separate boxplots are acceptable as parallel plots providing the scales for the two plots are the same so that comparisons can be easily made.

Part (b) is essentially correct (E) if teachers are picked based on some form of variability (e.g., range, compactness, IQR, spread, min-to-max, etc.).

Part (b) is partially correct (P) if either "Teachers" are selected with an incorrect explanation or no explanation, OR "Students" are selected based on center.

Note: If Ql = -0.5 and Q3 = 0.5 are being used for the students, then a choice of students based on IQR OR a choice of teachers based on range is scored essentially correct (E). An argument for students based on center stills scores as partially correct (P). A choice of one group over the other without indicating which measure of variability has been used is incorrect (I), but considering both issues without making a choice is an example of parallel solutions - each of which is essentially correct (E).

Part (c) is essentially correct (E) if the response says that testing these hypotheses will not answer the question posed and a plausible explanation is given. Plausible explanations should focus on things such as: positives and negatives canceling out, individual times versus group average, or variability. Reference to the boxplots is not expected.

Part (c) is partially correct (P) if the answer given is "No" and either there is no explanation and part (b) was scored as essentially correct (E),

OR

There is an incomplete or incorrect explanation that addresses the connection (or lack of connection) between the given hypotheses and the teacher's question. [Note: Answers that discuss conditions needed to perform the test do not address the issue of connection. Such responses are scored as incorrect (I), as are responses that attempt to change the teacher's question.]

Copyright © 2003 by College Entrance Examination Board. All rights reserved.

Available at apcentral.collegeboard.com,

3

Ap® STATISTICS

2003 SCORING GUIDELINE'S

Question 1 (cont'd)

4 Complete Response (3E)

All three parts essentially correct

3 Substantial Response (2E lP)

Two parts essentially correct and 1 part partially correct

2 Developing Response (2E OP or lE 2P or 3P)

Two parts essentially correct and no parts partially correct OR

One part essentially correct and 2 parts partially correct OR

two parts partially correct

1 Minimal Response (IE IP or lE OP or OE 2P)

One part essentially correct and either 0 or 1 parts partially correct OR

No parts essentially correct and 2 parts partially correct

Copyright © 2003 by College Entrance Examination Board. All rights reserved.

Available at apcentral.collegeboard.com,

4

Ap® STATISTICS

2004 SCORING GUIDELINES

Question 6

Solution

Part (a):

Step 1: States and checks appropriate conditions.

We are told that the sample was randomly selected. Since the sample size is large

(e.g., n =50 > 30), the one sample t interval should be valid. Alternatively, we could assume that the reduction in cholesterol level after one month is (at least approximately) normally distributed, but we have no way to check this assumption with the information provided.

Step 2: Identifies the appropriate confidence interval by name or formula.

One sample t interval for u, the mean reduction in cholesterol for the new drug or x ± (I . f- . "n

Step 3: Correct mechanics.

Degrees of freedom = 11-1 = 49.

~ * s 15 .

x±tn_1 r=24±2.0096 ~=24±4.2631=(19.7369,28.2631).

-sn ,,50

Step 4: Interprets the confidence interval in context.

We are 95% confident that the mean reduction in cholesterol for the new drug in the population of people with high cholesterol is between 19.74 and 28.26 mg/dl.

Part (b):

The decision based on a 95% confidence interval only corresponds to the two-sided test of significance at the 5% level, not necessarily the one-sided test. The confidence interval in (a) is equivalent to testing Ho: Il = 20 against Ha: )..l "* 20. In this test, the tail probability would be doubled, and this two-sided pvalue, .066, is larger than .05, failing to reject the null hypothesis. However, in testing Hs: Il = 20 against Ha: u > 20, the one-sided p-value of .033 is small enough to reject Ho at the 5% level.

Alternatively, if we had compared the p-value of .033 to an alpha level of .025, the conclusions would match.

Copyright © 2004 by College Entrance Examination Board. All rights reserved.

Visit apcentral.collegeboard.com (for AP professionals) and www.collegeboard.comiapstudents (for AP students and parents).

21

Ap® STATISTICS

2004 SCORING GUIDELINES

Question 6 (cont'd.)

Part (c):

The critical value for the lower confidence bound is the 951h percentile (instead of the 97.51h percentile) of the t distribution with 49 degrees offreedom.

l = 1.676

and

L = 24 ~ 1.676· S. = 24 - 3.5553 = 20.4447 y50

Part (d):

Yes, the decision would change. Since the lower bound L is more than 20, the agency would now be convinced that fl is greater than 20 and the new drug is statistically significantly better than the current drug.

Scoring

Part (a) is scored as

Essentially correct if all four steps (check conditions, identify procedure, calculate interval, interpret interval) are correct. Each step is scored as correct or incorrect, no partial credit is given for the steps.

• Ok if only state "n is large" or if assume population distribution is normal with some justification/recognition that this is only an assumption.

• Student can use either t interval or z interval but needs to give name of procedure (state t or z somewhere) or provide critical value.

• A correct interpretation of the confidence level does not count for step 4. An incorrect interpretation of level prevents credit for correct interpretation of interval.

Partially correct if two or three steps are correct.

Incorrect if at most one step is correct.

Part (b) is scored as

Essentially correct if the student discussion includes:

1. the confidence interval is two-sided and the test is one-sided, and

2. a quantitative linkage between the procedures (e.g., doubling the p-value, halving the level of significance, using 90% as the confidence level [with clear connection to the a = .05 significance level]).

Copyright © 2004 by College Entrance Examination Board. All rights reserved.

Visit apcentral.collegeboard.com (for AP professionals) and www.collegeboard.com/apstudents (for AP students and parents).

22

Ap® STATISTICS

2004 SCORING GUIDELINES

Question 6 (cont'd.)

Partially correct if

• Student states only that the confidence interval is two-sided and the test of significance is one-sided,

• Student has the right concept but discussion is very poor.

Incorrect if

• Student solution is only a restatement of the conclusions of each procedure.

• Student solution is only a restatement that the results differ.

• Student solution states that a 90% confidence interval is shorter and would not include 20 (but gives no indication of how 90% relates to the level of significance of the test) or that they might not differ if a different level of ex was used (with no value stated).

• Student says confidence intervals and test are different procedures (e.g., vary in specificity) and lead to different conclusions.

• Student says one uses the t value and the other uses the z value.

Note: Students may refer to the duality between confidence intervals and hypothesis tests in their solutions, but they must describe this duality in order to receive credit.

Part (c) is scored as

Essentially correct if

1. a reasonable one-sided value of t* is given (± 1.676 or ± 1.684), and

2. L is calculated correctly using the value of t* provided (20.445 Or 20.428).

Partially correct if the student identifies an incorrect reasonable critical value (z*, two-sided t* but see note below, or wrong df) and then uses this value to calculate L.

Incorrect if the student identifies a nonsensical critical value (e.g., uses the test statistic, the p-value, ex, t*12, 20) and/or obtains a lower bound larger than the sample mean (unless it is from an incorrectly substituted one-sided t* value).

Part (d) is scored as

Essentially correct if

1. the student correctly justifies whether or not the conclusion has changed, and

2. the student makes a correct conclusion from the one-sided confidence interval and supports their conclusion (e.g., compares L to the value of 20, believes f.L 2: L, is 20 in the interval),

Partially correct if student gives only component 2 or has component 1 with weak justification,

Incorrect if an answer (yes or no) is provided with no reasonable explanation.

NOTE: If the student's work in prot (c) is merely a reworking of the lower endpoint in (a) and they obtain the same lower bound and comment that the calculation is the same, they will receive at most one point among parts (a), (c) and (d). "ACD Rule"

Copyright © 2004 by College Entrance Examination Board. All rights reserved.

Visit apcentral.collegeboard.com (for AP professionals) and www.collegeboard.com/apstudents (for AP students and parents).

23

Ap® STATISTICS

2004 SCORING GUIDELINES

Question 6 (cont'd.)

Each Essentially Correct response counts as 1 point, each Partially Correct response counts as ~ point.

4 Complete Response 4 points
3 Substantial Response 3 points
2 Developing Response 2 points
1 Minimal Response 1 point IF A PAPER IS BETWEEN TWO SCORES (FOR EXAMPLE, 2~ POINTS) USE A HOLISTIC APPROACH TO DETERMINE WHETHER TO SCORE UP OR DOWN DEPENDING ON THE STRENGTH OF THE RESPONSE AND COMMUNICATION.

Copyright © 2004 by College Entrance Examination Board. All rights reserved.

Visit apcentral.collegeboard.com (for AP professionals) and www.collegeboard.com/apstudents (for Ai' students and parents).

24