Sie sind auf Seite 1von 63

Strategic Behavior in Individual Response

James Li Egan McComb Edmund Xu Sherry Zhang

10 December, 2010
Abstract

In this report, we analyze the expectations of strategic behavior among


students at Los Altos High School, specifically through the use of game
theory puzzles. Using a survey and various methods of statistical analysis,
we obtain both holistic and stratified results. We find that while students are
attuned to the behavior of the majority, they cannot effectively eschew this
behavior themselves to avoid the fallacy of composition. Furthermore, we find
that the correct minority response is the same as the majority response on an
equivalent but antithetical question. Lastly, we determine an answer to the
classic “2/3 of the average” game theory puzzle in a real world environment
which reproduces independent results.
Acknowledgments

We would like to thank the following teachers for their coöperation:

Lisa Bonanno
Teresa Dunlap
Carol Evans
Susan Friedeberg
Susana Herrera
Laraine Ignacio
Michael Moul
Perla Pasallo
Michael Richardson
Katherine Robertson
Galen Rosenberg
Michael Smith
Gabriel Stewart
Judy Strauss
Betty Yamasaki
Table of Contents
1 Introduction 1

2 Procedure 2
2.1 Survey Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Sample Design . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Administration . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Data Handling . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Data Summary 7

4 Results for Unstratified Data 8


4.1 Analysis of Responses to Questions 1–4 . . . . . . . . . . . . 8
4.1.1 Holistic Analysis of Responses to Question 1 . . . . . 9
4.1.2 Holistic Analysis of Responses to Question 2 . . . . . 10
4.1.3 Holistic Analysis of Responses to Question 3 . . . . . 11
4.1.4 Holistic Analysis of Responses to Question 4 . . . . . 12
4.2 Holistic Analysis of Response to Question 5 . . . . . . . . . . 14
4.3 Analysis of Responses to Questions 6 & 7 . . . . . . . . . . . 16
4.3.1 Holistic Analysis of Responses to Question 6 . . . . . 16
4.3.2 Holistic Analysis of Responses to Question 7 . . . . . 17

5 Results for Stratified Data 18


5.1 1–4 Stratified by Response to Related Question . . . . . . . . 18
5.1.1 Response to Question 4 vs. Response to 1 . . . . . . . 19
5.1.2 Response to Question 3 vs. Response to 2 . . . . . . . 21
5.2 Response to 5 Stratified In Various Manners . . . . . . . . . . 26
5.2.1 Response to Question 5 vs. Response to 7 . . . . . . . 27
5.2.2 Response to Question 5 vs. Inferred Comprehension . 29

6 Conclusion 35
6.1 Problems with the Sample . . . . . . . . . . . . . . . . . . . . 35
6.2 Problems with the Survey . . . . . . . . . . . . . . . . . . . . 36
6.3 Other Sources of Response Bias . . . . . . . . . . . . . . . . . 37
6.3.1 Contamination . . . . . . . . . . . . . . . . . . . . . . 37
6.3.2 Sincerity . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.4 Further Possibilities for Analysis . . . . . . . . . . . . . . . . 39
6.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

i
List of Figures
1 Unstratified Relative Frequencies — 1 . . . . . . . . . . . . . 9
2 Unstratified Relative Frequencies — 2 . . . . . . . . . . . . . 10
3 Unstratified Relative Frequencies — 3 . . . . . . . . . . . . . 11
4 Unstratified Relative Frequencies — 4 . . . . . . . . . . . . . 12
5 Unstratified Frequencies — 5 . . . . . . . . . . . . . . . . . . 14
6 Unstratified Boxplot — 5 . . . . . . . . . . . . . . . . . . . . 15
7 Relative Frequencies — 6 . . . . . . . . . . . . . . . . . . . . 16
8 Relative Frequencies — 7 . . . . . . . . . . . . . . . . . . . . 17
9 Relative Frequencies — 4 vs. A on 1 . . . . . . . . . . . . . . 19
10 Relative Frequencies — 4 vs. B on 1 . . . . . . . . . . . . . . 20
11 Relative Frequencies — 3 vs. A on 2 . . . . . . . . . . . . . . 21
12 Relative Frequencies — 3 vs. B on 2 . . . . . . . . . . . . . . 22
13 Relative Frequencies — 3 vs. C on 2 . . . . . . . . . . . . . . 23
14 Relative Frequencies — 3 vs. D on 2 . . . . . . . . . . . . . . 24
15 Relative Frequencies — 3 vs. E on 2 . . . . . . . . . . . . . . 25
16 Frequencies — 5 Given Knowledge of Game Theory . . . . . 27
17 Boxplot — 5 Given Knowledge of Game Theory . . . . . . . . 28
18 Frequencies — 5 Given Reasonability of Response . . . . . . 29
19 Boxplot — 5 Given Reasonability of Response . . . . . . . . . 30
20 Frequencies — 5 Juxtaposing Previous Strata . . . . . . . . . 31
21 Boxplot — 5 Juxtaposing Previous Strata . . . . . . . . . . . 32
22 Frequencies — 5 Another Juxtaposition of Strata . . . . . . . 33
23 Boxplot — 5 Another Juxtaposition of Strata . . . . . . . . . 34

ii
List of Tables
1 Raw Data — Bonanno . . . . . . . . . . . . . . . . . . . . . . 41
2 Raw Data — Dunlap . . . . . . . . . . . . . . . . . . . . . . . 42
3 Raw Data — Evans . . . . . . . . . . . . . . . . . . . . . . . 43
4 Raw Data — Friedeberg . . . . . . . . . . . . . . . . . . . . . 44
5 Raw Data — Herrera . . . . . . . . . . . . . . . . . . . . . . . 45
6 Raw Data — Ignacio . . . . . . . . . . . . . . . . . . . . . . . 46
7 Raw Data — Moul . . . . . . . . . . . . . . . . . . . . . . . . 47
8 Raw Data — Richardson . . . . . . . . . . . . . . . . . . . . . 48
9 Raw Data — Robertson . . . . . . . . . . . . . . . . . . . . . 49
10 Raw Data — Rosenberg . . . . . . . . . . . . . . . . . . . . . 50
11 Raw Data — Smith . . . . . . . . . . . . . . . . . . . . . . . 51
12 Raw Data — Stewart . . . . . . . . . . . . . . . . . . . . . . 52
13 Raw Data — Strauss . . . . . . . . . . . . . . . . . . . . . . . 53
14 Raw Data — Yamsaki . . . . . . . . . . . . . . . . . . . . . . 54
15 Master Schedule — English Courses . . . . . . . . . . . . . . 55
16 Master Schedule — Mathematics Courses . . . . . . . . . . . 56

iii
1 INTRODUCTION 1

1 Introduction
Game theory, a branch of applied mathematics, is the study of behavior in
strategic games wherein an individual’s favorable course of action is dependent
on those of the other equally rational players. Many such interactions involve
the unintuitive manifestation of the fallacy of composition, in which the
seemingly most advantageous course of action is disadvantageous for that
exact reason. For example, in the classic game of the prisoner’s dilemma,
two criminals want to minimize their jail time. Separated, they are offered
the opportunity to walk by defecting and betraying their accomplice since
the police have insufficient evidence to prosecute either of them for anything
other than a minor charge. Although it seems logical to defect and thereby
minimize jail time, it is actually highly undesirable because if each does so,
the police have sufficient evidence to prosecute them both. If neither defects
then the police have no choice but to charge them with nothing other than
the minor offense. The only win-win situation is to remain silent, but each
has to trust that the other will not betray him for a win-lose situation, and
in the real world players will usually gravitate towards the lose-lose situation
due to their suspicions of the other player.
Though often applied to social sciences like economics or politics, routine
life is abounding in such interactions, where the individual’s choice has an
effect on the efficacy thereof, or where the individual must preëmpt external
behavior. In this report, we will attempt to analyze the behavior of teenagers
at Los Altos High School when confronted with these situations through a
survey, specifically using non-coöperative, simultaneous n-person games, i.e.
games wherein players cannot form agreements, all the moves happen at
once, and there are an arbitrary, finite number of players. Whether or not
individuals in our sample will fall prey to the fallacy of composition is of
particular interest. For this reason, we will pose unembellished questions
both concerning the straightforward and the complex: the individual’s
expectations of the group’s strategy, and the individual’s evasive strategy
considering their expectations of those of the group.
To measure our population’s aptitude for predicting the behavior of each
other without the complication of the fallacy of composition, we will inquire
about their expectations of the majority response to the respective question.
To measure the cunning of our population at avoiding the behavior of each
other, we will inquire about their expectations of minority response to the
respective question. Lastly, we hope to obtain an answer to the “2/3 of the
average” puzzle in a real world environment.
2 PROCEDURE 2

2 Procedure
2.1 Survey Design
Designing surveys is particularly problematic due to the inefficacy of human
language when communicating with a diverse population. Firstly, the survey
questions must be communicated succinctly and precisely. The survey writer
must pose the question in such a way that the respondent can understand
precisely what is being asked, yet with no extraneous information that may
bias or confuse him. This is a difficult task indeed, for the imprecision
of natural language may impart a multitude of minutia or nuances that
are regarded differently from one individual to another. Unfortunately,
even the best possibly phrased questions can be ineffective because even
if no ambiguities exist, the response is still subject to the comprehension
of the respondent. In diverse populations, the survey writer can never be
sure of the literacy or acumen of her respondents. Thus a compromise in
communication must be found at the expense of the precision and succinctness
of thereof, maximizing public comprehensibility, unambiguity, and coverage
of contingency. This is the reason documents intended for use across large
populations are often verbose, repetitious, or insipid.
With our survey, we gather both primary data for analysis and secondary
data to be used as heuristics. The first four questions are multiple choice,
asking the respondents their expectations of the other respondents’ responses
to each respective question. The survey is prefaced with some simple instruc-
tions outlining how and where to write their responses as well as some less
important information regarding the structure of the survey and the rules
surrounding it, where the latter are enforced by the instructor, outlined in
their instructions.1 For formatting, font weights are modulated to indicate
emphasis. Of course, to be cordial the survey also includes a statement of
gratitude, even if it has little or no effect on the quality of the responses.
Questions 1 and 4 inquire about expectations of the majority response,
and questions 2 and 3 do just the opposite, i.e. expectations of the minority
response. Perhaps unusual questions, they are phrased to be as clear as
possible that they refer to themselves and not to some external set of choices.
The names of the possible responses are letters, with the first response
named ‘A’ and further responses named in alphabetical sequence. These
responses are not unique across questions because it is unknown what biases
respondents may have towards certain letters, and the responses were not
described in any way for the same reason. Each question is communicated
1
See 2.3 (p6) for more information on the administration of the survey.
2 PROCEDURE 3

identically but for the qualifying nouns “most” or “least” in order to minimize
the effect of linguistic nuance. The fifth question is free response, likewise
asking the respondents their expectations of the other respondents’ responses.
A much more technical question, and as such more difficult to communicate,
question 5 asks the classic “2/3 of the average” game theory puzzle. It is
problematic due to the use of mathematical terms that individual respondents
may not understand. We pose the question in a parallel manner as the other
four primary questions but unfortunately due to its nature we are forced to
use precise but perhaps esoteric terms such as “real number” and “mean” to
avoid ambiguity. Additionally, being a free response question the collection
of this data will be slightly more troublesome than the others.
Question 6, another multiple choice response, is included as a sincerity
heuristic, though the insincere respondent could just as easily lie on this
question as on any other; moreover the responses are highly subjective
and as such cannot be considered too seriously. Question 7 poses a very
straightforward, easy to communicate question, being whether or not the
respondent has heard of game theory, a simple Boolean response. Lastly,
question 8 attempts to assess the respondents’ confidence in their answers to
the five primary questions. This question is put quite plainly and formatted
with less finesse than the others.2 Included on the next page is a copy of the
survey administered.

2
See 6.2 (p36) for more information regarding its pitfalls.
2 PROCEDURE 4

School-Wide Survey

Please number and write your responses clearly on a separate piece of


paper. Each question is completely independent from the others but for the last three,
which are to gauge your responses. Refrain from conferring with your classmates, and
please answer all of the questions. You will have no more than five minutes.

1. Which choice of the below do you think the most people taking this survey will choose?

A)
B)

2. Which choice of the below do you think the most people taking this survey will choose?

A)
B)
C)
D)
E)

3. Which choice of the below do you think the least people taking this survey will choose?

A)
B)
C)
D)
E)

4. Which choice of the below do you think the least people taking this survey will choose?

A)
B)

2
5. Write a real number between 0 and 100 that you think will be closest to the mean
3
of everyone else’s response to this question.

6. How much thought did you give this survey?

A) None.
B) Some thought.
C) Much thought.

7. Have you heard of game theory?

Yes / No

8. Tell us whether you think you got each answer right:

1. Yes / No
2. Yes / No
3. Yes / No
4. Yes / No
5. Yes / No

Thank you for your response. It is greatly appreciated.


2 PROCEDURE 5

2.2 Sample Design


In order to obtain a good sample, we need to include students of all grade
levels and groups. The sample needs to be diverse, not just in age, but also
in academic level and background. Because we do not want respondents
conferring with each other, we choose to limit the survey to one period of one
day to minimize response bias. This limits the variety of classes from which
we can choose, since not all types of courses are available in even proportions
throughout the day.
First, we choose English courses because we can be reasonably sure that
these classes provide the greatest representation of our overall population
since every student must take them for four years. Further, we choose
mathematics courses because it adds diversity of mindset to our responses,
despite being prone to under-representation. Since our survey concerns
logical, analytical concepts, we suspect that students in a math class, with
a quantitative mindset, will respond slightly differently than students in
English class, with a qualitative mindset. Science and history courses provide
the same general dichotomy, but having the constraint of maximizing the
academic diversity of our sample in one period, we do not include them
to avoid further convolution. Elective courses cannot be included since
they over-represent a certain population and are not distributed evenly and
equivalently throughout the day.
With the constraints of administering the survey in one period, and to
only math and English courses, we analyze the master schedule to identify the
period with the most classes and the most diversity thereof.3 We determine
that second period has the most classes with greatest possible variety, and
thus choose to administer the survey therein. Our sample design is not
without its flaws, which will be analyzed in Section 6.1 (p35).

3
See Appendix B (p55).
2 PROCEDURE 6

2.3 Administration
We elect to conduct the survey on Wednesday. We hope that teachers are
more likely to oblige because the longer periods provide more class time.
Furthermore, the Wednesday selected was the first after the Veterans Day
four day weekend, so we can be reasonably sure that teachers do not plan to
administer an examination that would preclude extraneous activities, which
is common in long periods. After designing our sample and the logistics of
obtaining it, we contact the instructors of the selected courses via email,
inquiring whether they will be willing to present our survey to their second
period class.
Due to time constraints, besides communication by email, we contact
each instructor in person to give them an envelope with the necessary
materials. Included is a physical copy of the survey and some instructions
for the teachers. Teachers administer the survey for exactly five minutes,
and students are not allowed to collaborate with one another. The survey is
designed to be projected by a document camera, but if teachers so request,
digital copies are provided as well as class sets of physical copies in one
case. For the majority of the classes, students provide their own paper to
record their responses. After conducting the survey, teachers collect and
deposit the responses in the provided envelope, thereon writing the number
of students enrolled and the number of those present that day, so that we
may count individual nonresponse that would otherwise be undetectable. We
then collect the envelopes from each teacher later that day and proceed to
compile and analyze our results.

2.4 Data Handling


We manually digitize the surveys into non-headed, tab-delimited, ascii plain
text files for ease of data analysis. The surveys are blocked into separate files
by class, whose file name is in the format teacher.class.dat. From there,
the data is collated and filtered by standard Unix text processing utilities.
The tables included in this document are produced by shell scripts, with
minimal user intervention. Data blocking and stratification is implemented
in the awk language, interpreted by gnu awk 3.1.8. Statistical analysis
and data visualization are implemented in the s language, interpreted by r
2.12.0.
3 DATA SUMMARY 7

3 Data Summary
For a complete listing of the data collected, refer to Appendix A (p41).

• Classes Requested: 19

• Classes Sampled:4 14

• Individuals Enrolled in Sample: 363

• Individuals Attended in Sample: 336

• Individuals Responded in Sample: 334

• Total Question Nonresponse: 86

• Total Question Nonresponse, excluding #8:5 39

4
Refer to 6.1 (p35) for more information regarding total nonresponse.
5
Due to problems, #8 was thrown out. Refer to 6.2 (p36) for more information.
4 RESULTS FOR UNSTRATIFIED DATA 8

4 Results for Unstratified Data


We will begin by analyzing the collected responses holistically before pro-
ceeding to do stratifications thereon. The responses of interest in our survey
were the first five questions, with the remaining questions intended to serve
as heuristics by which to stratify the data, though at least the responses to
question seven provide an interesting statistic.

4.1 Analysis of Responses to Questions 1–4


The first four questions were multiple choice, and as such we classify responses
thereto as categorical variables. Categorical variables can be displayed in
various ways with various graphical representations, but we elect to show the
relative frequencies6 in a bar chart. The rationale for the choice of relative
frequencies is the fact that the number of observations for each question differ
slightly due to nonresponse, so plain frequencies would not have conveyed
the results as well. Use of bar charts is advantageous because the relative
magnitude of the scores can be easily ascertained qualitatively, though a
representation such as a pie chart would serve the same end.

6
The relative frequency is the score divided by the total number of observations.
4 RESULTS FOR UNSTRATIFIED DATA 9

4.1.1 Holistic Analysis of Responses to Question 1

Figure 1: Unstratified Relative Frequencies — 1

• Number of Responses: 332

• Number of Nonresponse: 2

• Correct Answer: A

Quite clearly, a significant majority of the responses to this question was ‘A’.
It seems some internalized analog of the ballot order effect caused respondents
to favor the first choice, whether rationalized or not. This mechanism was
obviously not completely prevalent however, as the frequency of the response
‘B’ was not vastly lower.
4 RESULTS FOR UNSTRATIFIED DATA 10

4.1.2 Holistic Analysis of Responses to Question 2

Figure 2: Unstratified Relative Frequencies — 2

• Number of Responses: 332

• Number of Nonresponse: 2

• Correct Answer: C

The results for this question are perhaps the most intriguing held herein.
Though the response ‘A’ is still quite prevalent, the overwhelming majority
of responses to this question was ‘C’, being 53.31% of the observations. In
this question the mantra, “when in doubt, choose C,” is quite apt. The
frequency of the other responses decline as they get further away from ‘A’,
though the relative frequency of ‘D’ is slightly higher than that of ‘B’.
4 RESULTS FOR UNSTRATIFIED DATA 11

4.1.3 Holistic Analysis of Responses to Question 3

Figure 3: Unstratified Relative Frequencies — 3

• Number of Responses: 330

• Number of Nonresponse: 4

• Fallacious Answer: E

• Correct Answer: C

Another intriguing result, these data clearly show the fallacy of composition.
The overwhelming majority of respondents chose ‘E’, at 44.24% of the
observations. In so doing, respondents ensured that it was the incorrect
answer. Interestingly enough, the correct answer was ‘C’ at 8.79% of the
observations, the same answer respondents chose for their answer to question
2, which is logically antithetical. This result makes sense logically, since
4 RESULTS FOR UNSTRATIFIED DATA 12

respondents are probably unlikely to choose the same choice for the majority
as the minority.7

4.1.4 Holistic Analysis of Responses to Question 4

Figure 4: Unstratified Relative Frequencies — 4

• Number of Responses: 328

• Number of Nonresponse: 6

• Fallacious Answer: B

• Correct Answer: A

Unlike the previous question, the fallacy of composition is much less apparent
in this instance since there are fewer choices. Furthermore, the results are
7
See 5.1.2 (p21) for elaboration.
4 RESULTS FOR UNSTRATIFIED DATA 13

much less conclusive, with the relative frequencies of each response being
essentially the same. However, it is important to note that the correct answer,
‘A’, is the same as the correct answer on the antithetical question 1, though
this may be coincidental.8

8
See 5.1.1 (p19) for more analysis.
4 RESULTS FOR UNSTRATIFIED DATA 14

4.2 Holistic Analysis of Response to Question 5


The fifth question was free response, and its variable a continuous, quanti-
tative one. We elect to display plain frequencies in a histogram so as not
to obscure the data. Histograms are advantageous because they show the
shape of a distribution plainly, and they hide a certain amount of information
that a stem and leaf plot does not, which is necessary since some responses
collected are whole numbers, other decimals, and others fractions whose
decimal equivalent is repeating. Additionally, we choose to display the data
in a box and whisker plot to easily communicate the minimums, maximums,
and quartiles visually and to complement the histogram.

Figure 5: Unstratified Frequencies — 5

The data display a right skew, though there is a peak at the left end of
the distribution. Responses were observed in the entire designated range,
throughout 0–100, as stipulated in the question.
4 RESULTS FOR UNSTRATIFIED DATA 15

Figure 6: Unstratified Boxplot — 5

• Number of Responses: 314

• Number of Nonresponse: 20

• Minimum: 0

• First Quartile: 20

• Median: 37.25

• Third Quartile: 54

• Maximum: 100

• Mean:9 38.50519

• Correct Answer: 25.67012

• People Correct:10 0

It seems that owing to the complexity of the question and the faults in its
communication that a large proportion of the respondents did not understand
it.11 Any response higher than 66 is impossible in all cases, for only in an
infinite population whose individuals all irrationally choose 100 would 66 be
the correct answer. Fortunately the third quartile is well within this limit,
though the first quartile is much higher than expected. Unmodified, these
data are not conclusive since they are occluded by the confounding variable
of comprehension. In Section 5.2 (p26) we attempt to obtain more conclusive
results using various stratifications.
9
We intend the arithmetic mean throughout, i.e. x̄ = n1 n
P
i=1 xi .
10
People whose response when rounded is equal to the rounded answer.
11
See 6.2 (p36) for elaboration on problems with the survey.
4 RESULTS FOR UNSTRATIFIED DATA 16

4.3 Analysis of Responses to Questions 6 & 7


For completeness, we present the results for questions 6 and 7. Like the first
four questions, we elect to display the relative frequencies of the respective
responses to these questions with bar charts.

4.3.1 Holistic Analysis of Responses to Question 6

Figure 7: Relative Frequencies — 6

• Number of Responses: 330

• Number of Nonresponse: 4

The majority of respondents indicated that they ”gave some thought” when
completing the survey. A highly subjective question, it is expected that
the neutral position be the most frequent. It is also noteworthy that more
4 RESULTS FOR UNSTRATIFIED DATA 17

respondents indicated that they ”gave little thought” to the survey than
”gave much thought”.12

4.3.2 Holistic Analysis of Responses to Question 7

Figure 8: Relative Frequencies — 7

• Number of Responses: 333

• Number of Nonresponse: 1

The majority of respondents had heard of game theory, though not over-
whelmingly so.

12
See 6.3 (p37) for more information on sincerity.
5 RESULTS FOR STRATIFIED DATA 18

5 Results for Stratified Data


In order to obtain more conclusive, detailed results, we will proceed to stratify
the data in various ways. More specifically, we will stratify the responses to
the first four questions by the responses to other questions in the primary
set, and we will attempt to unobfuscate the results to the fifth question using
responses to the heuristics questions as well as other metrics.

5.1 1–4 Stratified by Response to Related Question


Two types of questions comprise the first four questions: queries regarding
the expectations of majority and minority strategy respectively. Each of
these types are presented twice, one wherein there are but two response
choices, and one wherein there are five. We will examine these dichotomies
by comparing the responses to the antithetical questions in both of the latter
categories. Thus we will inspect the responses to question 4 versus those to
question 1, and those to question 3 versus those to question 2. We will use
the same graphical representations of the data as in the previous section.
5 RESULTS FOR STRATIFIED DATA 19

5.1.1 Response to Question 4 vs. Response to 1

Figure 9: Relative Frequencies — 4 vs. A on 1

• Number of Individuals in Stratum: 189

The vast majority of individuals who chose ‘A’ on the first question chose
the opposite response on the fourth question. As proposed before, it is clear
that individuals tended to choose the opposite response when differentiating
between their expectations of the behavior of the majority versus that of the
minority. This argument becomes more compelling when one examines the
other stratification of these data, in the next subsection.
5 RESULTS FOR STRATIFIED DATA 20

Figure 10: Relative Frequencies — 4 vs. B on 1

• Number of Individuals in Stratum: 139

Like in the previous stratification, these data indicate that the vast majority
of individuals who chose ‘B’ on the first question chose the opposite response
on the fourth question. Ironically, the fact that so many individuals chose
the opposite response makes it more intelligent to choose the same response
in both questions, however irrational that may seem. Indeed, in both of the
question regarding the minority, the correct answer is the same response
as the correct answer on the antithetical question, that which regards the
majority.
5 RESULTS FOR STRATIFIED DATA 21

5.1.2 Response to Question 3 vs. Response to 2

Figure 11: Relative Frequencies — 3 vs. A on 2

• Number of Individuals in Stratum: 61

Quite a bit more obscured than in the dual response questions 1 & 4, this
stratum does not clearly show a similar trend. It does offer a few insights,
however. The most frequent response was ‘E’, which happens to be the
fallacious, majority answer to the question itself. More interesting is that
larger proportions of individuals chose the responses ‘A’ and ‘C’ than did
choose ‘B’ and ‘D’, which are quite rational choices knowing the tendencies
of the sample’s expectancy of the majority response, being ‘A’ and ‘C’
respectively, and the fact that individuals may tend not to chose the same
response for antithetical questions.
5 RESULTS FOR STRATIFIED DATA 22

Figure 12: Relative Frequencies — 3 vs. B on 2

• Number of Individuals in Stratum: 36

Again, individuals in this stratum tended to favor the fallacious response ‘E’.
It is noteworthy that so small a proportion chose ‘C’ as their response, the
correct answer to the question.
5 RESULTS FOR STRATIFIED DATA 23

Figure 13: Relative Frequencies — 3 vs. C on 2

• Number of Individuals in Stratum: 177

In this stratum, the largest minority response was again ‘E’. A fairly large
proportion of individuals herein chose ‘A’, a thoroughly rational choice for
its seeming irrationality. It is again noteworthy that the lowest frequency
response was ‘C’, indicating a similar choice mechanism as seen in the
stratified dual response results.
5 RESULTS FOR STRATIFIED DATA 24

Figure 14: Relative Frequencies — 3 vs. D on 2

• Number of Individuals in Stratum: 36

Consisting of very few individuals, the analysis of this data should be regarded
with apprehension. Unlike the previous strata, the response ‘E’ does not
account for as high a proportion of the responses. Furthermore, the dearth
of ‘D’ responses again corroborate the theory postulated through analysis of
the dual response questions.
5 RESULTS FOR STRATIFIED DATA 25

Figure 15: Relative Frequencies — 3 vs. E on 2

• Number of Individuals in Stratum: 20

Consisting of even fewer individuals than the previous stratum, statistics


derived hereof are not to be considered very powerful. The proportions of
each observed response are roughly equal; the largest minority was response
‘D’ and there was an absence of response ‘C’, the correct answer to the
question.
5 RESULTS FOR STRATIFIED DATA 26

5.2 Response to 5 Stratified In Various Manners


Due to the complexity of the question and the obscured, inconclusive nature
of the results, we will examine the data for question five stratified in various
manners, including the use of the heuristics questions 6 and 7, and by
“inferred comprehension.”
For the heuristics questions, we will include individuals that responded
‘B’ or ‘C’ on question 6, i.e. those who gave “some” or “much” thought to
their responses, and individuals that responded in the affirmative to question
7, i.e. those who had heard of game theory. Our method for inferring
comprehension is simplistic and imperfect. We remove all responses 66 or
greater because they are impossible; only for an infinite population whose
individuals all irrationally choose 100 would 66 be the correct answer and
anything above that is a mathematical infeasibility. Unfortunately we cannot
justify removing similarly irrationally high responses on empirical grounds.
Doing so without empirical grounds is effectively data cherry picking, a
corrupt practice.
5 RESULTS FOR STRATIFIED DATA 27

5.2.1 Response to Question 5 vs. Response to 7

Figure 16: Frequencies — 5 Given Knowledge of Game Theory

The above histogram displays the distribution of responses to question


5 given that respondents affirmed knowledge of game theory in question 7.
This stratification evens out frequencies in the lower end of the distribution,
and lowers frequencies in the upper. Unfortunately it does not remove all
of the ludicrous responses, indicating that many individuals who had heard
of game theory either did not comprehend the question or did not think
sufficiently to obtain a more rational guess.
5 RESULTS FOR STRATIFIED DATA 28

Figure 17: Boxplot — 5 Given Knowledge of Game Theory

• Number of Individuals in Stratum: 170

• Minimum: 0.66

• First Quartile: 20

• Median: 33

• Third Quartile: 50

• Maximum: 88

• Mean: 36.18429

• Correct Answer: 24.12262

• People Correct: 4

As compared with the unstratified distribution, this box plot displays a


clear right skew with a lower median and third quartile. Furthermore the
interquartile range13 changes sufficiently to display the response 100 as an
outlier14 , which it certainly should be. The mean and thus the answer are
also lower than in the unstratified data.

13
The interquartile range, or IQR, is the difference between the third and first quartiles.
14
An outlier is defined to be 1.5×IQR outside the first and third quartiles, respectively.
5 RESULTS FOR STRATIFIED DATA 29

5.2.2 Response to Question 5 vs. Inferred Comprehension

Figure 18: Frequencies — 5 Given Reasonability of Response

Removing responses greater than or equal to 66 has a very straightforward


effect on the distribution. The frequency of responses in the range 0–10 is
relatively high, which is followed by a low proportion of responses in the
range 10–20 slowly climbing to a high proportion of responses in the range
40–50, falling precipitously thereafter as responses become more far-fetched.
5 RESULTS FOR STRATIFIED DATA 30

Figure 19: Boxplot — 5 Given Reasonability of Response

• Number of Individuals in Stratum: 256

• Minimum: 0

• First Quartile: 17

• Median: 33

• Third Quartile: 49

• Maximum: 65

• Mean: 30.99855

• Correct Answer: 20.66549

• People Correct: 13

The boxplot for this stratum shows the distribution is more symmetrical
than that of the unstratified data. Each of the metrics that comprise the five
number summary15 as well as the mean and answer are all of course lower.

15
The minimum, first quartile, median, third quartile, and maximum of a distribution.
5 RESULTS FOR STRATIFIED DATA 31

Figure 20: Frequencies — 5 Juxtaposing Previous Strata

Juxtaposing the previous strata, i.e. the knowledge of game theory and
the reasonability of response, we get the above distribution. This distribution
appears strikingly Gaussian but for spikes in the ranges 0–5 and 45–50.
5 RESULTS FOR STRATIFIED DATA 32

Figure 21: Boxplot — 5 Juxtaposing Previous Strata

• Number of Individuals in Stratum: 145

• Minimum: 0.66

• First Quartile: 17

• Median: 30

• Third Quartile: 44

• Maximum: 60

• Mean: 29.7264

• Correct Answer: 19.81758

• People Correct: 13

Just as in the previous stratum, the boxplot shows that the distribution is
fairly symmetrical. Furthermore, the tails are drawn in, and the quartiles are
slightly lower. This stratum allows the lowest mean and answer for question
5 yet.
5 RESULTS FOR STRATIFIED DATA 33

Figure 22: Frequencies — 5 Another Juxtaposition of Strata

Juxtaposing the last stratum, being the juxtaposition of knowledge of


game theory and the reasonability of response, with the favorable responses
to question 6 provides the above distribution. Overall, the shape is equivalent
to that of the previous stratum, though the frequencies are on average lower.
This shows that the sincerity of the response had little effect on the magnitude
thereof for this question.
5 RESULTS FOR STRATIFIED DATA 34

Figure 23: Boxplot — 5 Another Juxtaposition of Strata

• Number of Individuals in Stratum: 129

• Minimum: 0.66

• First Quartile: 17

• Median: 30

• Third Quartile: 43

• Maximum: 60

• Mean: 29.37813

• Correct Answer: 19.58542

• People Correct: 13

The boxplot indicates a distribution roughly equivalent to that of the previous


stratum, though the quartiles are lower. These data provide us the lowest,
least obfuscated mean and answer to question 5 of any of the stratifications
thereof, at the expense of some sample size which diminishes the favorable
effect of the Law of Large Numbers16 .

16
As the sample gets larger, the statistics converge on the parameter of the population.
6 CONCLUSION 35

6 Conclusion
6.1 Problems with the Sample
One major problem with our sample was the preponderance of nonresponse.
Of the nineteen classes in our sample, five classes did not respond in their
entirety. The reasons for this were varied: Ms. Arriada declined outright,
Mr. Chaffee forgot to administer the survey, Mr. Kanda’s class declined his
request, Ms. Robertson reneged at the behest of her class, and Mr. Spiteri
wasn’t able to respond to our request in time. Each class of nonresponse
causes sample bias. The exclusion of classes made our sample under- or over-
represent certain groups of our population. The effective sample, although
diverse, did not have equal proportions of different grades or academic levels,
which may have biased our data.
There were ways in which we could have improved our sample. Random
selection of individuals would have been ideal, eliminating sample bias.
Unfortunately, contacting and tracking down each individual would have
been highly impractical and inefficient. A stratified random sample would
have been still better, ensuring equal proportions of representation as in
the whole population. This method would have been inconvenient if not
impossible because we might have been barred from obtaining a list of all
students detailing what classes they were taking.
6 CONCLUSION 36

6.2 Problems with the Survey


The design of our survey had several flaws that potentially biased the
responses collected. To minimize the amount of paper we needed to print
in administering the survey to 19 classes of about 20–30 students each, we
required that students provide their own paper to complete the survey. This
made data collection and the transcription from paper to computer files
more difficult. Although we were quite careful in transcribing the students’
responses, it is possible that responses were miswritten or misread.
The fact that the survey was administered in most cases using a document
camera or a digital projector was also troublesome. Because the font was a
little small, some students had trouble reading the survey; Mr. Moul, whose
classroom is bigger than most, noted this in his comments. Furthermore,
the entire document could not fit on the screen at once and thus impeded
its administration since teachers had to move it as students completed their
responses, thereby forcing some students to wait before moving onto later
problems in order for others to complete the ones currently visible. Allowing
disproportional amounts of time on each question could easily have induced
bias in responses across classes.
Besides administration choices, the design of the questions themselves
was poor. The wording of question 5 was highly confusing to many students.
It can be clearly seen from the many instances of misguided responses that
the students did not understand what the question was asking. Thus their
answer could be a complete guess, which taints the results of that question.
Moreover, the wording and formatting of question 8 could have been much
better. Many students did not comprehend how to format their responses,
or to what the question referred, and as such many did not answer all five
parts thereof.
To ameliorate these problems, there are several changes we could make.
First, to make data collection easier, we could have given personal copies
to the students, or even have used machine-readable sheets in conjunction
therewith to eliminate problems potentially induced by human transcription.
For question 5, we could have given an example of what kind of answer we
are looking for, or how to determine an answer, e.g. “If you think the average
of all the responses to this question will be 33, then you should put 22 as
your response.” For question 8, we could have separated the constituents
and interlaced them among questions 1–5, and we could have been more
explicit in asking their confidence level, e.g. “Are you confident about your
answer to question 1?”. Each of these changes would improve the quality of
our data and eliminate some forms of bias.
6 CONCLUSION 37

6.3 Other Sources of Response Bias


6.3.1 Contamination
Response bias can also be brought about by contamination, either through
respondents answering questions with prior knowledge thereof, or through
respondents conferring with others taking the survey. These phenomena
can lead to results that do not accurately reflect the respondent’s true
understanding of the questions at the time of the survey’s administration.
We attempted to limit contamination by providing teachers with an
instruction sheet that emphasized that students were not to confer with each
other. We also established a universal time limit of 5 minutes, so that each
respondent would have the same amount of time to consider and complete
the survey. In addition, we scheduled all of the surveys to be administered
on the same day during second period. We hoped that these precautions
would effectively reduce the possibility of contamination to zero, though
contamination resulting from possible teacher negligence was still possible.
Our inability to administer the survey ourselves was perhaps the biggest flaw
in our prevention of contamination. Unfortunately, it is impossible to know
a posteriori whether the survey was administered correctly.
Judging by the number of nonresponse from teachers, it is reasonable to
conclude that the teachers who did administer the surveys were probably
not fully engaged in ensuring that the students were not conferring amongst
themselves or peeking at each others’ answers. While we did get confirmation
from the teachers that they did monitor the class according to the instructions
given, there is still reasonable suspicion based on the results of each class that
some classes may have had considerable breaches of secrecy. For example,
Mr. Stewart’s Global Connections class overwhelmingly answered ‘B’ to the
first question, despite the fact that the overall trend of the data collected
indicated that students favored ‘A’ instead. Additionally, most of the same
answers appeared consecutively when transcribed, which suggests that the
hypothetical groups who conferred with each other to produce a majority ‘B’
response likewise finished the survey at the same time and turned them in
together. This is purely conjecture, for there is no assurance that the order
of the responses in the collected envelopes had any relation to the order in
which they was collected, and by extension to the hypothetical groups that
colluded to obtain their answers. In addition to larger scale contamination,
there was inevitably much micro-contamination that would be even harder
to locate.
6 CONCLUSION 38

6.3.2 Sincerity
Sincerity on the part of the respondents is very important, since insincerity
on any level would negatively impact the efficacy of our results. As shown
by the responses to question 6, most respondents gave “some thought” to
the survey, followed by “little thought”, and lastly “much thought”. Since
respondents are often biased towards selecting answers based on their beliefs
of what the questioners want, it is very likely that a large portion of the
majority who chose “some thought” actually gave the survey little or no
thought at all. In hindsight, this question would probably have been far
more accurate had we given the respondents but two choices to choose from:
“some thought”, and “no thought”, since these disparate choices would force
them to reveal their sincerity with out being masked by subjectivity.
One of the biggest factors supporting the suspicion that many of our
respondents were indeed insincere is rooted in the demographics of our
sample population, consisting entirely of teenagers, known for their general
apathy. The lack of incentives, and perhaps the vapidity of the survey do
nothing to help this situation. One possible improvement would again be to
administer the surveys in person, since our population would be more likely
to answer sincerely when confronted with peers in earnest than an austere
survey forced upon them by their teacher. Unfortunately, due to logistics
this was impossible.
6 CONCLUSION 39

6.4 Further Possibilities for Analysis


Because of the comprehensive nature of our data, having taken great pains
to include and record different groups of respondents either by their ethea or
their responses to the survey itself, there exist plenty of further stratifications
than those that we analyzed. For example, we could have examined the
data stratified by class, grade, or academic level. These three aspects of the
individual respondent could easily be significant, because grade level can
explain the maturity of the respondents, and academic level indicates to
a certain extent their mental capability. Stratifying by class, on the other
hand, would be helpful in locating the biases within, including the possibility
of contamination or insincerity.
We had attempted to stratify the sample population by confidence through
our position of question 8. Unfortunately, we had to nix this plan due to
the large volume of nonresponse, which we determined was the result of
poor wording and formatting on our part. If we had successfully posed the
question, however, we would have been able analyze the frequency of correct
answers versus the respondents’ confidence thereof to see if confidence had
any bearing at all on the verity of the responses.
Another very interesting opportunity for analysis would have been to
infer the thought patterns for successive weakly-dominated strategies to
question 5. For example, we could first eliminate every guess that is 66.66 or
above because of its infeasibility. If we eliminate these responses, then any
guess that is 44.44 or above is weakly-dominated for every respondent. This
elimination method would continue iteratively until the Nash Equilibrium
is reached at a response of 0. We could then stratify the responses based
on these strategy ranges to glean some important information about the
respondents’ mastery of the game.
6 CONCLUSION 40

6.5 Summary
Despite various shortfalls in the administration and collection of data, there
are several conclusions that can be made. First is that the response ‘A’ is
popular, being the majority response when given choices of ‘A’ and ‘B’, and
being the second highest frequency response when given the choices ‘A’–‘E’.
Second is the intriguing verity that the response ‘C’ is the overwhelmingly
most frequent response when given the choices ‘A’–‘E’. As a follow-up, it
would be interesting to see if these results are reproducible in different
character ranges, and if some similar analog exists for numbers or other
enumerable items.
Third, most students do not realize the fallacy of composition, as shown
by the fact that the response ‘E’ out of ‘A’–‘E’ was so overwhelmingly
popular. We cannot discount the supposition that the individuals in our
sample reasoned that ‘E’ was most likely to be the least popular answer,
and thus the most likely chosen for the minority response. Knowing this, ‘E’
becomes the most likely least likely to be the minority response, and then
most likely not chosen therefore, making it the most logical choice. It is much
more probable however, that these individuals did not realize any of that at
all, and instead answered simplistically, ignoring the fallacy of composition.
Which brings us to our fourth conclusion, that individuals were unlikely to
select the same answer for their expectations of the minority strategy as they
did for their expectations of the majority strategy, since as we saw in both
questions 3 and 4 that the correct answer ended up being the very same as
that to the antithetical questions 1 and 2, respectively.
Lastly, we determined an answer to the “2/3 of the average” game theory
puzzle in a real world setting, though confounded by problems of compre-
hension or sincerity. Obtaining an answer of 19.8 after data manipulation
and 25.7 without, it is interesting to note that our values are close to those
obtained in other experiments, where Wharton economics majors produced
26.6, Caltech general undergraduates produced 20.0, and a public sample of
19,196 Danes for a prize of 5000kr obtained 21.6.17 18 Despite our worries that
the performance of Los Altos students was poor since the answer produced
was so far from the Nash Equilibrium of 0, it seems that even highly educated
people fall prey to the same logical limitations.

17
Nagel, Rosemarie (1995). “Unraveling in Guessing Games: An Experimental Study,”
American Economic Review 85, 1313-1326.
18
Schou, Astrid (2005). “Gæt-et-tal konkurrence afslører at vi er irrationelle,” Politiken.
A Collected Data Tabulated by Class
Each table is composed of records representing one survey collected and fields
representing the responses therein. The fields correspond to each respective
question on the survey, and are encoded as follows:

• Fractional responses to question five are converted to decimal numbers.

• Boolean responses are indicated with ‘1’ or ‘0’ for ‘yes’ or ‘no’ respec-
tively.

• Responses to number eight are conglomerated.

• Nonresponse on a question is indicated with a dash (‘-’).

Table 1: Raw Data — Bonanno


Bonanno — Survey of Composition and Literature
A A E B 0.2 B 0 00000
A E E A 47.5 A 0 11111
A C E B 60 B 0 00000
B E B A 60 B 0 11000
A A B B 25 B 0 10010
A C D B 57 B 0 00111
A C E B 37 B 0 11111
A C A B 24 A 1 11110
A C E B 66 B 1 11111
B C D A 66 A 1 11110
A C E B 34 C 0 11110
A C B A 1 A 0 11111
A C E B 10 A 0 11111
B C A A - A 0 00000
A A B A 0.6 B 0 11110
A A B B - B 0 00000
B C E A 60 B 0 11010

41
Table 2: Raw Data — Dunlap
Dunlap — Geometry Honors
A A A A 50 A 0 00000
A A A A 88 B 0 11101
A A D B 20 C 1 11010
B B E A 50 B 1 11010
A C E B 65 C 0 10001
B C E - 33 C 1 11000
A A E B 63 C 0 11110
A A D B 40 A 0 11000
A C E B 77 B 0 11110
B D E A 72 B 0 11110
A D E B 30 B 0 10100
A C E A 36 C 0 11000
A C C A 24 B 0 01011
A C B A 20 B 0 10101
B E D A 22 B 1 00000
B D A A 30 C 0 00100
A C C A 22 C 0 11110
A A C B 88 B 0 11010
A C E B 66 B 0 10010
A C E B 37 A 0 10110
A C E B 35 B 1 11110
A C C B 15 B 0 00000
A A B B 66 B 0 10110

42
Table 3: Raw Data — Evans
Evans — Algebra II Honors
B C A A 30 B 1 10010
A C E B 25 B 1 10010
A D B B 44 B 1 10101
A C B B 20 C 1 01101
A A D B 0.66 C 1 11111
A C A A 20 B 1 11111
B D A A 72 B 1 10010
B D A A 33 C 1 00010
A C C B 5 B 1 10100
A A E B 30 B 1 11000
A C D B 36.1 B 1 10001
A A A A 2 A 1 11110
A C A B 67 B 0 10011
A C E A 27 C 1 11110
B C E A 15 B 1 11110
A C E B 44 C 1 11111
B C D A 33 B 1 11010
B C E A 21 B 1 11100
B D C A 20 B 1 11110
B B E A 30 C 1 11111
A A B B 1 B 1 11111
B E D A 16 B 1 00000

43
Table 4: Raw Data — Friedeberg
Friedeberg — Algebra II
B B C A 2 B 0 00000
A A - - 66 C 1 00110
B B D A 22 C 1 11110
A A E B 1 B 0 11110
A B E B 1 C 1 10101
B C E A 28 B 1 11111
A C B B 69 B 1 11111
B C E A 20 B 1 11010
A C A B 15 B 1 01111
B B E A 15 B 0 00000
A C D B 25 B 1 11100
B B E A 1 C 1 00100
B B E A 50 B 0 11111
A A C B 5 B 1 -----
A C E B 50 B 1 11111
A A C B 1 B 1 00000
B C E A 25 B 1 11110
A D E B 50 C 1 11111
A A C B - B 1 00000
A C D B 33 B 1 11001
B C D A 16 B 1 01110
A C C B 7 B 0 00000
B D A B 1 B 1 -1- - -
A C E B 50 B 0 01000
A C E B 40 C 1 10110

44
Table 5: Raw Data — Herrera
Herrera — World Literature Honors
A C A A 13 A 0 00100
A B E B 69 A 1 00000
B C D A 4 B 0 00000
A C D B 74 B 0 11010
B D C A 42 B 0 10100
A C E B 54 C 1 11010
B C A B 17 A 0 00000
A C B B 75 B 1 11011
A B E B 22 B 1 11101
B D A A 57 C 1 10010
B C E A 3 B 1 01100
A A A A 30 B 1 11000
A C D B 30 B 0 00000
B C E A 50 B 0 00100
A C E B 57 B 1 10110
A C B B 50 B 0 10111
A C E B 75 B 1 11100
A C E B 50 B 0 10110
B D E A 75 B 1 10010
B B A A 40 C 1 11110
B C E A 14 C 1 11011
B E E B 14 B 1 00001
B A C A 43 B 1 11100
A C D B 35 B 1 10010
B A C A 72 B 1 11100
A C D B 35 B 0 10111

45
Table 6: Raw Data — Ignacio
Ignacio — Algebra I
B C E A 30 B 0 11111
B C A A 65 C 0 11111
B C A A 70 B 0 11001
A C A B 66 B 0 11010
A C E B 50 A 0 00000
B C E A 23 A 0 11111
A E E A 3 A 0 00000
B B A A 50 B 0 11001
A E E B 50 B 0 11011
- - - - 25 A 0 00000
B C A A - B 0 11111
B C D A 50 B 0 11110
A A C B 33 A 0 11111
B C C A 50 B 1 01000
B D B A 77 A 0 11111
A A D B 22 B 1 10010
A C D B 20 A 0 11010
B C E A 69 A 1 00000

46
Table 7: Raw Data — Moul
Moul — Language and Composition AP
A C E B 50 B 1 01101
A C D B 24 B 0 10010
A C E B 52 B 0 11010
B B B B 11 B 1 11111
B C E A 37.5 B 1 01100
A C B B 23 B 1 00001
B C A A 57 B 1 10010
B E B A 33.3 C 1 11111
A C D - 50 A 1 11010
A C C B 22 B 0 11111
B D C A 8 C 0 00000
A C E B 3.6 B 1 11011
A C E B 7 B 0 00000
A A E B 0 C 0 11111
A A E B 46 A 0 00000
A C B B 66 B 0 00000
B C A A 50 B 0 00000
A B E B 50 B 1 10011
A B B B 33.3 B 1 11111
B E A A 50 B 1 11011
A A A A 50 A 1 11111
A C D B 66 B 0 11111
B C D A 28 B 1 00100
A A E B 50 B 1 11110
A C A B 50 A 1 11111
A C E B 67 B 1 10100
B C E B 9 A 0 10000
A A E B 50 A 0 01010
B C E A 50 A 1 10000
A D E B 32 B 0 11111

47
Table 8: Raw Data — Richardson
Richardson — Calculus BC AP
A A A A 25 B 1 11110
A A A A 100 A 1 11110
A D E B 52 B 1 11110
A C D B 77 B 1 11000
B B E A 18 B 1 11111
B D B A 10 B 1 11110
A C E A 40 B 1 11110
B C E B 1 B 1 11111
A D A A 43 B 1 10010
A B E B 20 B 1 11110
A C E B 50 B 1 10110
A C E A 33 B 1 11111
A A D B - A 1 11111
A A E B 20 B 1 00000
A C D B 33 B 1 00000
B C A A 33 B 1 11010
B C A A 52 B 1 11111
B C E A 37 B 0 11110
A D E A 46 B 1 11011
B C E A 36 B 1 10110
B D A A 42 C 1 11110
A C E B 60 B 1 11010
A C E A 50 B 1 11010
B B D A 33 B 1 11100
B D E A 15 B 1 11111
A C D A 25 B 1 00000
A A C A 33 C 1 11110

48
Table 9: Raw Data — Robertson
Robertson — Algebra II
B D E A 69 B 1 11111
B B D A 69 C 1 10101
A C D B 54 B 1 11011
B C E A 69 B 1 11111
A C E B 2 B 1 01100
A C D B 32 B 1 10010
B C A A 20 B 1 11111
A A E B 66 A 1 11111
B C A B 69 C 1 11111
A C D B - A 1 11111
A A E B 1 A 1 00000
A A D B 7 B 1 11110
B D A A 50 B 1 11110
B D A A 32 B 1 11001
B C E A 75 B 1 11100
B C E B 55.5 A 1 11111
B C E A 5 A 1 00000
B C E A 88 B 1 10110
A C E B 50 B 1 11110
A C E B 25 C 1 11110
B D E A 56 A 1 11110
B C A A 50 B 1 11000
B C D A 35 B 1 11111
B B A A 13 A 1 10101
B C E A 50 B 1 10110
B C A A 46 B 1 -----
B D A A 66 B 1 11111

49
Table 10: Raw Data — Rosenberg
Rosenberg — American Literature
A C E B 32 B 0 11110
B C E A 5 B 1 11110
A E A A 69 A 1 11111
B C E A 40 A 0 10100
B B E A 75 B 0 10110
A C A B 42 B 1 10110
A C E B 35 B 1 01100
A E D B 66 B 1 11010
B C E A 40 B 0 11100
B B E A 27 C 1 00000
B C A A 7 B 1 -----
A C E B 33 A 1 11111
B B E A 35 C 0 10001
B E A A 60 B 0 10010
A A E B 35 A 0 11110
B D A A 46 B 0 10101
A C A B 55 A 0 -----
B C E A 40 A 0 10110
A B D B 7 B 0 11010
A B E B 40 B 0 10100
A C B B 17 B 1 11110
A E D B 66 A 0 10101
B C A A 40 B 0 00000
A A A A 20 A 1 11111
B B B B 40 B 0 11111
B B A A 40 B 0 11111
B C A A 34 A 0 11111

50
Table 11: Raw Data — Smith
Smith — Creative Views
A C A B 75 A 0 10101
B C E A 74 A 0 11010
A C E B 50 C 1 11110
B C E A 27 B 0 11111
B E A A 50 B 1 11111
B B D A 60 C 1 11111
A E D B 20 - 0 11110
A A E B 75 C 0 11110
A E B B 80 B 0 11001
A D E B 70 B 0 11000
B C E A 50 B 0 11000
B C A A 75 A 0 11111
B C D A 20 B 1 10011
B C E A 36 B 0 11110
B A E A 50 B 0 10101
A C A B 75 C 0 10010
A C D B 75 B 0 11110
B E A A 8 B 0 10010
A C E B 60 B 0 11110
A E B B 3 B 1 1111-
A C E B 60 B 1 011- -

51
Table 12: Raw Data — Stewart
Stewart — Global Connections B
A A E B 50 B 0 11111
B C A A 5 B 1 11110
B C A A 50 B 0 00000
B C D A 46 A 0 00000
B C E A 32 C 0 11111
B E D A 69 B 0 10101
B C E A 14 B 1 11010
B C E A 50 B 0 00000
B B A B - B 1 11111
A A E B 10 B 1 11110
B B E A 69 A 0 11111
B B E A 3 B 0 11110
A C A B 48 A 0 00000
B D C B 48 B 0 00000
B C E A 6 C 0 -----
B C E A - A 0 10000
A C E B 22 A 0 01100
A C D B 44 B 0 11011
A A E B 10 B 1 00000
B C C B 4 C 0 11111
B E B A 10 A 0 11111
B D E A 24 B 0 11111
A A A A 0 A 0 00000
A A D B - A 0 00000
B C D A 66 B 0 01011
A C E B 66 C 0 11110
A A D B 1 B 0 11110
A A E B 69 A 1 11111
A A E B - B 0 10101
A A E B - B 0 10011

52
Table 13: Raw Data — Strauss
Strauss — Trigonometry Honors
B D A A 66 B 1 10101
A A A A 0.66 B 1 11111
A B E B 42 A 1 10110
B C A A 30 B 1 11010
A C C A 40 B 0 00000
A A C B 50 C 1 00000
A C C B 19 A 1 01100
A B C B 63 B 0 11101
B C B A 44 C 1 01111
A A E B 42 B 1 11110
A C E B 41 B 1 11111
A A E B 37 B 0 11111
A B D A 36 B 0 10000
A A A A 50 A 1 00000
A A D A 10 B 0 11000
A A C B 66 B 1 10010
B C D A 66 B 0 10111
B B B A 11 B 0 0110-
B C A B 7 C 1 00010
A C D B 17 B 0 11111
B C A A 27 B 1 11010
A C B B 30 B 1 11111
A D E B 40 C 0 10010

53
Table 14: Raw Data — Yamsaki
Yamsaki — Calculus
A A C B 10 B 1 00000
A C E B 42 B 0 11110
A C D B 30 B 1 11010
B C E A - B 1 11110
B C D A - A 1 11110
B D E A 66 B 1 10010
A C E B - A 1 -1- - -
B B E A - A 1 10010
A D C B 70 B 0 10011
A D - - - - 1 -----
A D A B - B 1 10110
A D C B 75 A 0 00000
A C D B 28.6 C 1 11011
A A A A 50 A 0 00000
- - - - - A - —–
A A E B - - 0 11110
A C E B - - 1 11110
A A B A 30 B 0 10110

54
B Master Schedule of Courses to Sampled

Table 15: Master Schedule — English Courses


Courses Available
Period

American Lit

AP Language
AP Literature
Creative Views
Global A
Global B

Film Analysis

Survey Comp/Lit

Survey Skills

World Lit

World Lit H
0
1 3 3 3 3 33 3
2 3 3 3 3 3 3 3 3
3 3 3 3 3 3 3
4 3 3 3 3 3
5 3 3 3 3 33 3
6 3 3 3 3 33
7 3 3 3

55
Trigonometry H

3
3

3
Trigonometry

3
3
3
3
3
3
3
Math Mastery

3
3
Math Lab
Table 16: Master Schedule — Mathematics Courses

Geometry H

3
3
3
3
3
3
3
Geometry Alt. Path

3
Geometry 9
3
3
3
3
3
3
3
Geometry
3

3
Courses Available

3
Calculus BC
3

3
Calculus AB
3
3

3
3
Calculus
CAHSEE Math 3
3

AP Statistics
Algebra II H
3
3

33 3

Algebra II
3
3
3
3
3

Algebra I Enhanced
3

3
3

Algebra I
3

3
3
3

Algebra 9
3

3
3
3

Period
0
1
2
3
4
5
6
7

56

Das könnte Ihnen auch gefallen