pol 138 exam 3
Political scientists weight survey data for which ONE of the following reasons?
Because the population characteristics did not match the sample characteristics
Suppose that, on a test in a given school, mean scores are as follows: 90 for freshmen, 60 for sophomores, 60 for juniors, and 80 for seniors. In a linear regression predicting scores on the test using only predictors for "freshman" (coded 1 for freshman and 0 otherwise), "sophomore" (coded 1 for sophomores and 0 otherwise), and "senior" (coded 1 for seniors and 0 otherwise), what would be the coefficient on "senior"?
20
The probability of X happening is 50%, and the probability of Y happening is 50%. X and Y are independent events. What is the probability that X and Y both occur?
25%
Suppose that the 500 test scores for Group A follow a normal distribution and have a mean of 600 and a standard deviation of 100. About 95% of these test scores should fall between which two scores?
400 and 800
Suppose that we conduct an N>100 experiment to assess whether watching the news causes people to be more likely to vote. Suppose that, in reality, watching the news does not cause people to be more likely to vote. The probability that our experiment produces a p-value of 0.05 or less is ___.
5%
Suppose that men are 50% of a population, but men are 60% of our sample. If weighting sample respondents only on gender, the weight that should be applied to each man's response is ___.
50/60
Suppose that students in a class take a 100-item test. The mean number of correct items on the test is 60, and the standard deviation of the number of correct items is 8. If the teacher counts each item correct as 1 point and then adds 5 points to each student's score, the standard deviation of the test scores would be
8
Statistical power calculations indicate that there is an 80% chance of detecting a d=0.70 effect size at p=0.05 for a two-tailed test, if each sample has 33 observations. What would be the chance of detecting a d=0.70 effect size at p=0.05 for a two-tailed test, if each sample has 103 observations?
Greater than 80%
The t-statistic for a test is -3.0 in which the sample size is sufficiently large. Based on this, which of the following is true about the corresponding p-value?
The p-value is less than 0.05
A researcher conducted a randomized experiment and then compared the mean response from participants in the control to the mean response from participants in the treatment. The p-value for the difference in mean responses was p=0.99. Based on this p-value, the researcher should ___.
conclude that there is not enough evidence to conclude that there is an effect
Suppose that a sufficiently large set of students takes a new IQ test twice in immediate succession and that the correlation between the students' first score and the students' second score is 0.99 (p=0.01). Should that correlation be interpreted to be good evidence that this new IQ test has high validity?
no
Suppose that a school had two sets of students: Group A students, and Group B students. At this school, the number of Group A students equals the number of Group B students. For a test, members of Group A had a mean score of 50, and members of Group B had a mean score of 50. 1. Does this mean that members of Group A were just as likely to get a 50 on the test as members of Group B were? 2. Does this mean that members of Group A were just as likely to get a 70 on the test as members of Group B were?
no and no
For a null hypothesis of no difference, if the t-static for the test of a hypothesis is t=0.02, then we should do which of the following?
none of the above
If the p-value for the test of a hypothesis is p=0.50, then we should do which of the following?
none of the above
Suppose that a score of 60 is at the 40th percentile for scores on a test. This means that ___.
none of the above
Suppose that research indicates that men are more likely than women to support Candidate X (p<0.001). Based on this pattern, we should ___.
not conclude that gender influences support for Candidate X
For which of these 95% confidence intervals for a linear regression coefficient is the associated p-value less than 0.05? [Select all that apply and only those that apply]
o [1, 5] o [-5, -1]
Suppose that, for a set of participants randomly selected from a population, the 95% confidence interval for support for the president is [60, 70]. If we were to lose a random half of these responses and re-calculate the 95% confidence interval, then that new 95% confidence interval would be ___.
wider than [60, 70]
If an expert witness presents evidence from 10 studies, each with a power of 10% and each of which found statistically-significant evidence for the effect, that ___ suspicious.
would be
If an expert witness presents evidence from 3 studies, each with a power of 90% and none of which found statistically-significant evidence for the effect, that ___ suspicious.
would be
If an expert witness presents evidence from 4 studies, each with a power of 90% and each of which found statistically-significant evidence for the effect, that ___ suspicious.
would not be
If the p-value from a single statistical test of a null hypothesis is p=0.01, do we have enough evidence to claim that there is statistically-significant evidence for the detected association?
yes
Is a measuring device that is always off by 10% a reliable measuring device?
yes
Suppose that a sufficiently large set of students takes a new IQ test twice in immediate succession and that the correlation between the students' first score and the students' second score is -0.01 (p=0.01). Should that correlation be interpreted to be good evidence that this new IQ test has low reliability?
yes
Suppose that a sufficiently large set of students takes a new IQ test twice in immediate succession and that the correlation between the students' first score and the students' second score is -0.01 (p=0.90). Should that correlation be interpreted to be good evidence that this new IQ test has low validity?
yes
Suppose that you are representing plaintiffs suing ChemCorp, which produces Chemical X. Plaintiffs claim that Chemical X has caused the plaintiffs' leukemia. A research lab has conducted 20 research studies, each of which found a correlation between Chemical X and leukemia; however, none of the studies had a p-value under 0.05. These were the only 20 studies ever conducted by anyone. Is it in your clients' interest to combine the studies together into a meta-analysis?
yes
Which ONE of these t-statistics would have the lowest associated p-value, all else equal?
-10.00
Suppose that the mean height in a population is 180 cm and that the standard deviation of height in that population is 10 cm; suppose also that heights in this population follow a normal distribution. Candidate X is 160 cm tall. What is the z-score that is associated with Candidate X's height in that population?
-2
Bob receives a 20 on each of his 10 exams. What is the standard deviation of Bob's exam scores?
0
Which ONE of the following values indicates the correlation in the list that is the weakest correlation that can occur in real data?
0.0
If you flip a fair coin 4 times, which of the following is the correct way to calculate the probability of all four flips landing on heads?
0.5 * 0.5 * 0.5 * 0.5
If we flipped a coin 4 times and got 2 heads and 2 tails, what would the p-value be for a statistical test of the null hypothesis that the coin is fair?
1
Which ONE of these p-values is possible?
1
Which one of these is NOT a necessary step in a randomized experiment involving human participants?
After the difference in treatment, use control variables to eliminate alternate explanations.
suppose that I want to study whether Illinois residents support splitting Illinois into two states: Northern Illinois, and Southern Illinois. I mail a survey to all students at Illinois State University about this question. What is my study's population?
Illinois residents
Statistical power calculations indicate that there is an 80% chance of detecting a d=0.70 effect size at p=0.05 for a two-tailed test, if each sample has 33 observations. What would be the chance of detecting a d=0.20 effect size at p=0.05 for a two-tailed test, if each sample has 33 observations?
Less than 80%
The probability of A happening is 10%, and the probability of B happening is 40%. A and B are not independent events. What is the probability that A and B both occur?
None of the above or Cannot be determined from the information provided
Suppose that scores on a national test follow a normal distribution and have a mean of 600 and a standard deviation of 100. If Student A raises her score from 500 to 600, and Student B raises her score from 800 to 900, which ONE of the following statements is true?
Student A had a higher percentile increase on the test.
For which of the following sets of coin flips would the p-value be lowest for a test of the null hypothesis that the coin is fair?
TTTTTT
Suppose that 10 students take a test and that each student gets a different score on the test. The teacher then subtracts 10 from each student's test score. In this case, the correlation between students' first test score and students' test score after the score adjustment will be ___.
a positive correlation and will be a correlation of 1
Suppose that 100 students take a test and that each student gets a different score on the test. The teacher then added 5 to the score for only those students who scored below the mean score on the test. In this case, the correlation between students' first test score and students' test score after the score adjustment will be ___.
a positive correlation but not necessarily a correlation of 1
The set of numbers {⎼4, ⎼6, ⎼8} has a standard deviation that ___.
above 0
The phrase "post hoc ergo propter hoc", which refers to a logical fallacy, is best translated as___.
after this, therefore because of this
Suppose that a researcher is assessing whether Asian Americans are more likely or less likely to support Candidate X, compared to Americans who are not Asian American. The researcher analyzes data at the state level and discovers that states with higher percentages of residents who are Asian American have higher mean levels of support for Candidate X than states with lower percentages of residents who are Asian American. The researcher therefore concludes that Asian Americans are more likely to support Candidate X, compared to Americans who are not Asian American. Of the options below, this weakness in the research design is BEST described as ___.
an ecological fallacy
Suppose that you are asked to create a ten-item test that can be used to measure political interest. 1. Explain how the reliability of that measure could be assessed.
as long as the response relates to the idea that reliability concerns how consistent the measure is. For example: Check whether responses from participants are consistent over time.
Suppose that you are asked to create a ten-item test that can be used to measure political interest. 1. Explain how the validity of that measure could be assessed.
as long as the response relates to the idea that validity concerns whether the measure measures what the measure is supposed to measure. For example: Check whether persons with higher levels of political interest are more likely to watch TV news.
Suppose that the mean height in a population is 170 cm and that the standard deviation of height in that population is 10 cm; suppose also that heights in this population follow a normal distribution. Regarding height in this population, a candidate who is 180 cm tall would be in which of the following ranges?
between the 2.5th percentile and the 97.5th percentile
Of the following, imputation in political science is BEST described as ___.
using known information to make informed guesses about unknown information
Which one of the following regards whether a measure measures what the measure is supposed to measure?
validity
How do control variables help improve causal inference in a correlational study?
eliminate alternate explanations
It is a ___ to believe that there is no effect when there truly is an effect.
false negative
It is a ___ to believe that there is an effect when there truly is no effect.
false positive
Generally speaking, as p-values get lower, the associated t-statistics ___.
get farther from zero
All else equal, the larger the sample size, the ___.
higher statistical power is
Suppose that, in a population in which the number of men equals the number of women, women on average support Candidate X at 30% and men on average support Candidate X at 50%. If men were oversampled in our sample, then, without weighting the data, the mean support for Candidate X in the sample is more likely to be ___.
higher than 40%
Suppose that students in a class take a 100-item test. The mean number of correct items on the test is 60, and the standard deviation of the number of correct items is 8. If the teacher counts each item correct as 2 points, the standard deviation of the test scores would be ___.
higher than 8
In an experiment, Bob injected treatment lab rats with Chemical X; control rats were not injected with Chemical X. The data indicated that the treatment rats one week later had higher levels of skopetia than the control rats did. The treatment rats had been injected with 10cc of Chemical X but, when analyzing the data, Bob's assistant thought that the treatment rats had been injected with 1.0cc of Chemical X. Because of this error, the assistant's estimate for the influence of Chemical X on skopetia is likely to be ___.
higher than the true value
Suppose that a set of numbers has a mean of 20, a median of 20, and a standard deviation of 5. Suppose that the number 80 were added to this set of numbers. The addition of the number 80 to that set would ___.
increase the standard deviation of the set of numbers
All else equal, the larger the variation in a measurement, the ___.
lower statistical power is
All else equal, the smaller the association that a study is testing for, the ___.
lower statistical power is
1. Removing an outlier from a set of data would have more influence on the ___ of the data.
mean
Suppose that I predict student scores on a final exam for a sufficiently large set of students using student scores on their midterm. For some students, my prediction is too high; for other students, my prediction is too low; and, for some students, my prediction is exactly correct. The mean student score on the final exam is 70, and the mean of my predicted scores was 70, so ___.
my predictions had random error
Suppose that I predict student scores on a final exam for a sufficiently large set of students using student scores on their midterm. For some students, my prediction is too high; for other students, my prediction is too low; and, for some students, my prediction is exactly correct. The mean student score on the final exam is 70, and the mean of my predicted scores was 80, so ___.
my predictions had systematic error
Suppose that it is true that, the older a person is, the more likely it is that the person votes. Suppose that we have a variable Y, coded as the year that the participant in our study was born. Suppose that we have another variable V, which is coded 0 for a person who did not vote in the most recent election and coded 1 for a person who did vote in the most recent election. In this case, Y and V should be ___.
negatively correlated
If the p-value from a single statistical test of a null hypothesis is p=0.99, do we have enough evidence to claim that there is statistically-significant evidence for the detected association?
no
Suppose that a population has only two groups: A and B. Suppose that 20% of members of Group A are on public assistance and that 40% of the population are members of Group A. Does this mean that members of Group A are underrepresented among persons in the population who receive public assistance?
no
Suppose that a researcher tests whether taller candidates are more likely to be elected than shorter candidates are. The researcher analyzes voting data from the years 2000 through 2015 and finds that the taller candidate in an election has on average been more likely to win election than the shorter candidate in the election (p<0.01). Does this study have a high amount of internal validity?
no
Suppose that a sufficiently large set of students takes a new IQ test twice in immediate succession and that the correlation between the students' first score and the students' second score is -0.01 (p=0.90). Should that correlation be interpreted to be good evidence that this new IQ test has high reliability?
no
1. Standard deviation is a measure of ___.
variation
Suppose that, for a class of 101 students, a researcher codes each student's rank on the final exam from 1 (for the highest score) to 101 (for the lowest score). Each student received a final exam score that is a whole number between 0 and 100; no student received the same score as another student. 1. The correlation of the final exam scores with the final exam ranks could be ___. [Select all and only those options that apply. None might be correct.] 2. The correlation of the final exam scores with the final exam ranks will certainly be ___. [Select all and only those options that apply. None might be correct.] 3. The standard deviation of final exam scores could be ___ the standard deviation of final exam ranks. [Select all and only those options that apply. None might be correct.] 4. The standard deviation of final exam scores will certainly be ___ the standard deviation of final exam ranks. [Select all and only those options that apply. None might be correct.]
o negative o exactly -1 o negative o exactly -1 o the same as o the same as
Suppose that you have a meta-analysis with a large number of studies testing the same hypothesis drawing samples randomly from the same population. Some studies have a large sample size, some have a small sample size, and some have a medium sample size. The meta-analysis contains all studies conducted on this hypothesis. 1. What should be the correlation between sample size and standard errors among the studies in the meta-analysis? 2. What should be the correlation between sample size and effect sizes among the studies in the meta-analysis?
o negative correlation o no correlation
If Y decreases when X decreases, then that is a ___.
positive correlation
All else equal, adding participants to a study measuring support for the president should do what to the standard error for support for the president?
reduce the standard error
For a null hypothesis of no difference, if the p-value for the test of a hypothesis is p=0.02, then we should do which of the following?
reject the null hypothesis and accept the alternative hypothesis
Which one of the following regards the consistency of a measure?
reliability
Suppose that a researcher is interested in whether conservatives are more interested in the news than everyone else. The researcher attends a Donald Trump rally and surveys attendees. Results indicate that persons at the rally who identified as conservative were more interested in the news than the typical U.S. resident is. Of the options below, this weakness in the research design is BEST described as ___.
selection bias
Suppose that we are testing the null hypothesis that a coin is fair. All else equal, the p-value should be ___ for an unfair coin, compared to a fair coin, if the coin is flipped a large number of
smaller
If we flipped a coin 5 times and got 0 heads and 5 tails, what would the p-value be for a statistical test of the null hypothesis that the coin is fair?
something between 0 and 1
If, in an experiment, the mean for the control group was 2, and the mean for the treatment group was 3, what would the p-value be for a statistical test in which the null hypothesis is that the control group mean is equal to the treatment group mean?
something between 0 and 1
The term "statistically significant evidence" refers to ___.
sufficient evidence that a particular association exists
Which ONE of the following would be wider, for the same set of measurements?
the 60% confidence interval for the mean
Researchers not reporting studies that they conduct is referred to as ___.
the file drawer problem
In the past few years, there have been several unsuccessful attempts to get the same results as prior studies have gotten. This is referred to as ___.
the replication crisis
Test A has a mean of 50 points and a standard deviation of 10 points. Test B has a mean of 70 points and a standard deviation of 10 points. Both Test A and Test B have scores that follow a normal (bell curve) distribution. Student A scored at the 50th percentile on Test A, and Student B scored at the 50th percentile on Test B. Student A and Student B were then each given an extra 5 points on their test. This extra 5 points would have produced ___.
the same percentile increase for Student A on Test A as for Student B on Test B
Suppose that, for a set of participants randomly selected from a population, the 95% confidence interval for support for the president is [60, 70]. If we were randomly select a lot more participants from the population, ask these participants their support for the president, and then re-calculate the 95% confidence interval for support for the president, then that new 95% confidence interval would be ___.
thinner than [60, 70]
Suppose that, for a set of participants, the 95% confidence interval for support for the president is [60, 70]. The 90% confidence interval would be ___.
thinner than [60, 70]
Suppose that a researcher is interested in whether higher levels of income cause persons to be more likely or less likely to support Candidate X. The researcher finds that persons with higher levels of income are less likely to support Candidate X than persons with lower levels of income. The researcher therefore concludes that higher levels of income cause persons to be less likely to support Candidate X. Of the options below, this weakness in the research design is BEST described as ___.
under-controlling
Suppose that you conducted an experiment that involved randomly injecting laboratory rats with a substance in a vial marked "100% Chemical X" and comparing cancer rates in those rats to cancer rates for rats in a control group that were not injected with the substance in the vial. However, the vial was truly a 50% solution of Chemical X. This error will cause you to ___ the effect of Chemical X on cancer rates in rats, if Chemical X truly causes cancer in lab rats.
underestimate