Exam 3AB 2003

CPS questions (1-8)

1-3)             We wish to test the model that the probability of the ball landing on a red cell in a roulette wheel is ½.  For the following possible data, which consequences (A, B, C) apply? (MTF)

 

A) The data refute P = ½

B) The data are irrelevant to P = ½

C) The data support P = ½

 

1) (2 pts) Data:  20 spins, all Red                                   A    B    C

2) (2pts)  Data:  Payoff percentage at a slot machine     A    B    C

3) (2pts)  Data:  1000 spins, 523 Red                             A    B    C

 

4) (5 pts)  We considered the correlation that red cars have higher accident rates (color is variable X, accident rate is variable Y).  Which of the following causal models of this correlation use a “third variable” to explain the correlation?  (That is, for which models are X and Y not connected by a causal chain?)  MTF

 

Causal model

Cause

Third variable?

accident-prone drivers prefer red

 

Fill in (A) if  a 3rd variable is invoked

people become more aggressive when driving red

 

Fill in (B) if a 3rd variable is invoked

more dangerous cars tend to be painted red

 

Fill in (C) if a 3rd variable is invoked

red is harder to see, so a red car is more likely to be hit by other drivers

 

Fill in (D) if a 3rd variable is invoked

 

5) (5pts) Again consider the reason why red cars have higher accident rates (a correlation).  If “type of car” (not color) is the true cause of accident rates, which accident rate levels are expected in cells 1 & 2 of the following table?  (one only)

Car color,

accident rate

red

high

(1)

other

(2)

low

 

sports

“safe”

 

type of car

 

A)  1 is high, 2 is high

C)  1 is low, 2 is high

B)  1 is high, 2 is low

D)  1 is low, 2 is low

 

 

6) (5 pts) Which of the following options would be experiments, from the lecture on autism and Facilitated Communication (and the lecture that followed)?  These were some of the different ways we might try to find out if the words being typed were from the facilitator.  The question is merely which are experiments, not which would be effective.  MTF

A)    Watch closely and look for evidence that the Facilitator is controlling the typing.

B)     Change the positions of the letters on the keyboard and see if the words change.

C)     Show the child but not the Facilitator something and ask for a description of it.

D)    Monitor brain waves of child and Facilitator to determine who is responding.

7) (5pts) The test of FC shown in the video was an experiment because (one only)

A) it included controls

C) it manipulated the normal FC environment

B) it included replication

D) all of the above (A-C)

 

8) (4pts)  If a lab says that the DNA type of a crime scene sample matches the DNA type of a suspect, what does it mean if the random match probability (RMP) is 1/billion but the lab error rate is 1%?  In other words, how should a jury deal with those two numbers in assessing the significance of a match.  (one only)

A)    The significance calculation is not affected.

B)     The proper calculation is 1/billion x probability that no error was made (0.99)

C)     The proper calculation is 1/billion PLUS 1%

D)    You cannot calculate the significance of a match if there is a chance of human & technical error.

 

Criminal Justice System

 

9) (5pts) What evidence was discussed in class that reveals specific problems in the procedures used and the ways data have been gathered in the criminal justice system?  If the point wasn’t made, or if the option gives a substantially different impression than was made in class, do not mark that option.  MTF

A)    From dozens of wrongful convictions evaluated by the Innocence Project, mistaken identification, false confessions, and bad hair matching (with methods that did not use DNA) were factors contributing to the wrongful conviction in many of the cases.

B)     Although Federal and many State prosecutors are formally trained in the uses of scientific methodology (trained in scientific methods relevant to procedures used by courts), this training deviates from accepted methods used by most scientists.  

C)    Since the routine use of DNA, the prosecution’s prime suspect has been cleared before trial approximately 25% of the time, revealing that the prosecution’s methods had led them to the wrong person up to that point.

D)    For many of the scientific methods of analyzing data used in criminal trials (such as fingerprints, microscopic hair matching, and eyewitness identification) proficiency tests and/or experiments of these methods have revealed error rates of 10% or more.

 

 

10) (5pts) When couched in terms and concepts of our course, the major, widespread violations of ideal data evident in the criminal justice system included which of the following? MTF:

A)    The prosecution fails to gather and analyze data blindly, resulting in selective reinforcement of evidence against the prime suspect.

B)     Bad standards are common, including failures to conduct blind proficiency tests of the labs, and inadequate (sometimes non-existent) reference databases.

C)    Bad protocols.  In some cases, methods have been used widely despite the lack of protocols for analyzing the data.  In other cases, protocols for gathering the data have been inadequate for protecting against human and technical error.

 

Correlations

 

11. (5 pts) Which points about correlations are true? MTF

(A) The main reason correlation does not imply causation is that simple correlations are consistent with many causal models

(B) A zero correlation between X and Y means that there is no tendency for the average of Y to change as X changes.

(C) A correlation can exist when one or both variables are categorical (e.g., male or female), but it may not always be possible to indicate whether such a correlation is positive or negative.

(D) (Non-zero) correlations are of interest to society despite their limitations because so few are discovered.  Thus, when one is discovered, it is desirable to use it to the fullest.

 

 

12. (5 pts; hard) Although Western medicine has been highly successful at preventing many infectious diseases through vaccinations, it has not been successful at preventing cancer.  Instead, the major medical advances with cancer have been in early diagnosis, followed by treatment to kill or remove the cancer.  New diagnostic methods have allowed us to detect cancers at earlier stages of the disease, and the average age of a person first diagnosed with cancer has decreased.  In addition, people have been living longer from the times of their diagnosis. That is, if A is the average age of a person when cancer is diagnosed and Y is the number of years a person lives beyond diagnosis, there is a (negative) correlation between A and Y (Y increases as A decreases).  Which of the following models are consistent with this correlation (not rejected by the correlation)?  To think about this question properly for some options, you should realize that the average age of death of a diagnosed person is A+Y.

(A)  People who are diagnosed with cancer today tend to die at older ages than people diagnosed in the past.

(B)  People who are diagnosed with cancer today tend to die at younger ages than people diagnosed in the past.

(C)  Early diagnosis causes a person to live beyond the time they would die if diagnosed late

(D)    Early diagnosis causes a person to die before the time they would die if diagnosed late

 

 

 

13. (5 pts) Which options describe non-zero correlations (=any correlation that is defined but is not zero)? MTF

(A) People are more likely to catch influenza (the flu) in December than in June.

(B) The average length of a Bio301D lecture is the same during the first week of class as during the last week of class.

(C) People living in Ohio eat eggs for breakfast twice a week on average; people living in New Mexico play the lotto 3 times a week on average.  (This one is not easy, unless you think about it).

 

14-16. (4 points each) Researchers have observed that global carbon dioxide (CO2) levels in the atmosphere have increased in the last 100 years and that the earth’s average temperature has increased as well.  The increase in CO2 levels can be attributed to human burning of fossil fuels (oil, coal), and there is widespread concern now that the increase in atmospheric carbon dioxide is causing the increase in global temperature, through what is known as a “greenhouse effect.”

Use the following variables:                    Variable 1: time (100 years ago until now)

Variable 2: global CO2 level

Variable 3: global temperature

For each of the following questions, you are given a pair of these variables.  You are asked to choose among the following 3 options that best characterizes their relationship in the problem description above. 

 

(A) no correlation or causation is indicated.

(B) a correlation is indicated, but no causation between the variables is suggested

(C) a correlation is indicated and a causal relation between the variables is suggested.

For each pair of variables given below, which option applies (one answer each)?   

14.(4pts) Variables 1 & 2:        (A) (B) (C)

15.(4pts) Variables 1 & 3:        (A) (B) (C) 

16.(4pts) Variables 2 & 3:        (A) (B) (C)


 

Controls

17. (5 pts) Mark all of the following statements about controls that are correct.  MTF

(A)  Controls are present in any correlation, even if the correlation is zero.

(B)  To establish a correlation between smoking and lung cancer, at least some of the individuals must be smokers.

(C)  A control group needs to be chosen blindly to qualify as a true control

(D)  The choice of subjects randomly guarantees that a control is present.

(E)   By definition, controls are present in any experiment

 

 

18. (6 pts) Each of rows (A)-(F) describe different treatments that could be applied to humans in generating data on heart disease.  The treatments differ in which factors are present (indicated by “+”) or absent (-). Factor 1 is the use of a cholesterol-lowering drug (statins); factor 2 is a diet low in saturated fat; factor 3 is moderate exercise; factor 4 is a diet with daily fish oil; factor 5 is one ounce per day of alcohol.

Which two treatments would you want to compare to determine if factor 1 is correlated with differences in heart disease when all other factors are controlled? In evaluating possible answers, pick any comparison that controls for all unwanted factors, and assume that these six treatments differ only in the ways stated. Mark exactly two options, or none if none apply.  Each row (each option) describes a different set of conditions, so to know which factors would be applied in a treatment, you look across the row.  If multiple combinations satisfy the problem, any correct combination will be accepted.  (Two answers or None).

 

factor

 

Option

 

1

2

3

4

5

(A)

+

-

+

+

+

(B)

-

+

-

+

-

(C)

+

+

+

-

+

(D)

-

-

+

+

-

(E)

-

+

-

+

-

(F)

-

+

+

-

+

 

 

19. (6 pts) An epidemiologist does a survey to evaluate whether a child is more likely to develop leukemia when living in a house near a power-line transformer than when living in a house not located near a power-line transformer. The study uses all children from ages 5-15 in one city over a period of 10 years.  The study group includes boys and girls, different neighborhoods, from families with different income levels.  These different variables (age, gender, neighborhood, family income level) are recorded and, aside from whether the child develops leukemia, are the only data available for use in the analysis.  With the data obtained in this study, which of the following factors are or could possibly be controlled for in the analysis?  MTF

(A) child gender

(D) family income level

(B) neighborhood of residence

(E) child performance in school

(C) city of residence

 

 

 

 

 

 

Experiments

20. (5 pts). Which design features are typically included in experiments but lacking from correlation studies? (at least one answer)

A) Explicit protocols

C) Blind

E) None

B) Replication

D) Controls

 

 

21. (7 pts) Prisoners of Silence.  Which of the following options either depict what was shown in the video, or correctly explains the nature of the study done in the video?  If any part of an option is incorrect, consider the option incorrect and do not mark it.  MTF.

(A) Experiment: the reason that the tests shown are considered experiments is because they followed an explicit protocol and were carried out with the goal of determining whether the Facilitator or child was the source of the words.

(B) Replication: the video described multiple levels of replication of the results, not just the use of several pairs of pictures shown to the same facilitator pair

(C) Controls: the controls for the picture identification tests were the cases in which the child and facilitator were shown the same photo, and the correct word was typed.  This part of the design controls for all possible factors in the testing arena that might differ from the normal FC setting.  Thus when the Facilitator and child were shown different pictures, the outcome could be attributed only to the Facilitator not knowing what the child saw, and vice versa. 

(D) Blind: The design was “double blind,” meaning that the child did not see the picture shown to the Facilitator, and the Facilitator did not see the picture shown to the child.  The study would have been just as valid if the child was allowed to see both pictures, but the Facilitator was only shown the one picture (hence still blind), provided the same results had been obtained (that the words always described the picture seen by the Facilitator) .

22. (5 pts) This question refers to first two experiments shown in the "Secrets of the Psychics" video (horoscopes and palm reading) and to the “personality profile” experiment done with our class.  Which statements about design features are true?  Some options address all 3 experiments, some only one.  This question omits any reference to the Moscow picture psychics. 

A)    All 3 experiments clearly involved replication and blind, but all 3 experiments either did not include randomization or were ambiguous about randomization

B)     A control group was not explicitly (directly) shown for the horoscope experiment in the video.

C)     The following would constitute a control for our 301 personality profile experiment:  give all students the same personality description, but indicate on some forms that everyone is getting the same description as theirs.  This group would serve as a control in evaluating the effect of doing the experiment blind

D)    Randomization: given the way that the personality profile study was conducted in our class, with everyone receiving the same profile, randomization was not relevant in assigning the profiles to students.

 

23. (4pts)  Key code.  Fill in (A B) to indicate your key for this version of the exam.  Be sure your name and pad number are correctly bubbled in on the scantron and that you have signed this exam form.