BBH 310 exam 2 questions || Already Graded A+.
A drug doesn't work. But a pharmaceutical company keeps running studies different ways until
finally, one study gets a significant benefit of the drug. They hide the results of the non-
significant studies, and only publish the significant results for everyone to see. The result they
publish is a :
A. True positive finding, indicating that their way of doing research is both specific and sensitive
B. False positive finding, indicating that their way of doing research is not-specific enough to
only find results of drugs that really work
C. True negative finding, indicating that their way of doing research is both specific and
sensitive
D. False negative finding, indicating that their way of doing research is not-sensitive enough to
detect the effects that drugs really have correct answers B
If we set our alpha level (cutoff between significant and non-significant) at 0.05 and run 40
studies on a drug that doesn't work, how many false positive findings should we expect?
A. 0
B. 1
C. 2
D. 3
E. 4 or more correct answers C
Imagine we have a drug that does help lower blood sugar, but it only helps a little bit, and it
doesn't work for everyone. The researchers are scared of type 2 errors. They come up with a list
of things that they should do to reduce the risk of type 2 errors. Which reason does not belong on
the list, because it will actually increase the risk of type 2 error?
A. Get a very big sample size
B. Chose ways of measuring blood sugar with less variability (e.g. Hemoglobin-AIC instead of
fasting blood sugar)
C. Reduce the significance threshold alpha correct answers C
Which statistical test do you think would be appropriate for a clinical trial testing a drug against
a placebo, where the outcome of interest is the number of cigarettes smoked in the last week of
the trial?
A. Correlation
B. Chi-Squared
,C. T-test
D. ANOVA correct answers C
The t-statistic in a T-test approximately represents:
A. The probability that a result at least this unlikely appeared by random chance
B. How many standard errors apart are two means
C. The slope of the line of best fit for standardized variables correct answers B
The p value in a chi-squared test indicates the probability that under the null hypothesis, there
would be at least as much difference between:
A. The probability of being in certain categories of variable A depending on which categories of
variable B the subjects were in
B. The mean of variable A depending on which category of category B the subjects were in
C. The expected value of variable A depending on how high is variable B for each subject
correct answers A
Which of the following is correct?
A. Alpha was 0.01, p was 0.03, the null hypothesis can be rejected
B. Alpha was 0.01, p was 0.03, the null hypothesis cannot be rejected
C. Alpha was 0.05, p was 0.03, the null hypothesis cannot be rejected correct answers B
Review question: What's a standard deviation?
A. A measure of how spread out the variable is, in terms of typical difference between each data
point and the mean
B. A measure of how far a single data point is from the mean, in terms of difference from the
mean difference
C. A measure of the effect of sampling on the mean of the distribution for a given sample size,
showing how large of a difference we expect when taking another new sample and finding the
mean agai correct answers A
Which of the following describes a statistical type 1 error?
, A. Two students in BBH 310 each flip a different coin 50 times, then they do a statistical test to
see if there's an association between the flips one student got and the flip the other student got,
the resulting p value is 0.4971
B. Smoking truly does cause accelerated mortality rates, and a cohort study investigating
whether smoking is associated with higher mortality finds a p value for that test of 0.0098
C. A drug does not actually have any influence at all on a disease, but a researcher runs a clinical
trial testing for effects of that drug on that disease, and gets a p value of 0.0098
D. A drug truly is a little bit beneficial for treating a disease, but because of a small sample size,
a clinical trial testing this drug for this disease finds a p value for the effect of the drug on the
disease of 0.4971 correct answers C
A lab is designing a new diagnostic test for the detection of malaria. They are especially
concerned about correctly identifying everyone with malaria (not missing any actual cases). This
means they're especially concerned with not making _________ and value ________ over
_________ .
A. Type 2 errors; specificity; sensitivity
B. Type 2 errors; sensitivity; specificity
C. Type 1 errors; specificity; sensitivity
D. Type 1 errors; sensitivity; specificity correct answers B
Type 1 error risk goes up if
A. all of these increase type 1 error risk
B. you have a small sample
C. you have a small alpha
D. you try many tests correct answers D
When are researchers most likely to fail to find a significant correlation, even though the
variables involved are truly correlated in real life?
A. When the design is a cohort study
B. When alpha is relatively high
C. When sample size is relatively low
A drug doesn't work. But a pharmaceutical company keeps running studies different ways until
finally, one study gets a significant benefit of the drug. They hide the results of the non-
significant studies, and only publish the significant results for everyone to see. The result they
publish is a :
A. True positive finding, indicating that their way of doing research is both specific and sensitive
B. False positive finding, indicating that their way of doing research is not-specific enough to
only find results of drugs that really work
C. True negative finding, indicating that their way of doing research is both specific and
sensitive
D. False negative finding, indicating that their way of doing research is not-sensitive enough to
detect the effects that drugs really have correct answers B
If we set our alpha level (cutoff between significant and non-significant) at 0.05 and run 40
studies on a drug that doesn't work, how many false positive findings should we expect?
A. 0
B. 1
C. 2
D. 3
E. 4 or more correct answers C
Imagine we have a drug that does help lower blood sugar, but it only helps a little bit, and it
doesn't work for everyone. The researchers are scared of type 2 errors. They come up with a list
of things that they should do to reduce the risk of type 2 errors. Which reason does not belong on
the list, because it will actually increase the risk of type 2 error?
A. Get a very big sample size
B. Chose ways of measuring blood sugar with less variability (e.g. Hemoglobin-AIC instead of
fasting blood sugar)
C. Reduce the significance threshold alpha correct answers C
Which statistical test do you think would be appropriate for a clinical trial testing a drug against
a placebo, where the outcome of interest is the number of cigarettes smoked in the last week of
the trial?
A. Correlation
B. Chi-Squared
,C. T-test
D. ANOVA correct answers C
The t-statistic in a T-test approximately represents:
A. The probability that a result at least this unlikely appeared by random chance
B. How many standard errors apart are two means
C. The slope of the line of best fit for standardized variables correct answers B
The p value in a chi-squared test indicates the probability that under the null hypothesis, there
would be at least as much difference between:
A. The probability of being in certain categories of variable A depending on which categories of
variable B the subjects were in
B. The mean of variable A depending on which category of category B the subjects were in
C. The expected value of variable A depending on how high is variable B for each subject
correct answers A
Which of the following is correct?
A. Alpha was 0.01, p was 0.03, the null hypothesis can be rejected
B. Alpha was 0.01, p was 0.03, the null hypothesis cannot be rejected
C. Alpha was 0.05, p was 0.03, the null hypothesis cannot be rejected correct answers B
Review question: What's a standard deviation?
A. A measure of how spread out the variable is, in terms of typical difference between each data
point and the mean
B. A measure of how far a single data point is from the mean, in terms of difference from the
mean difference
C. A measure of the effect of sampling on the mean of the distribution for a given sample size,
showing how large of a difference we expect when taking another new sample and finding the
mean agai correct answers A
Which of the following describes a statistical type 1 error?
, A. Two students in BBH 310 each flip a different coin 50 times, then they do a statistical test to
see if there's an association between the flips one student got and the flip the other student got,
the resulting p value is 0.4971
B. Smoking truly does cause accelerated mortality rates, and a cohort study investigating
whether smoking is associated with higher mortality finds a p value for that test of 0.0098
C. A drug does not actually have any influence at all on a disease, but a researcher runs a clinical
trial testing for effects of that drug on that disease, and gets a p value of 0.0098
D. A drug truly is a little bit beneficial for treating a disease, but because of a small sample size,
a clinical trial testing this drug for this disease finds a p value for the effect of the drug on the
disease of 0.4971 correct answers C
A lab is designing a new diagnostic test for the detection of malaria. They are especially
concerned about correctly identifying everyone with malaria (not missing any actual cases). This
means they're especially concerned with not making _________ and value ________ over
_________ .
A. Type 2 errors; specificity; sensitivity
B. Type 2 errors; sensitivity; specificity
C. Type 1 errors; specificity; sensitivity
D. Type 1 errors; sensitivity; specificity correct answers B
Type 1 error risk goes up if
A. all of these increase type 1 error risk
B. you have a small sample
C. you have a small alpha
D. you try many tests correct answers D
When are researchers most likely to fail to find a significant correlation, even though the
variables involved are truly correlated in real life?
A. When the design is a cohort study
B. When alpha is relatively high
C. When sample size is relatively low
