Midterm 1 Practice Questions 1 KEY
These questions will help you prepare for Midterm 1. Some of them take the form of midterm questions (i.e, multiple-
choice questions with 5 answer options), while others are of a different format but focus on concepts you should be
familiar with.
1. A researcher completes a paired samples t test on 27 pairs of participants. If her observed t=2.033, which of the
following is a correct statement of the results?
a. t(27)=2.033, p<.05
b. t(27)=2.033, p>.05
c. t(26)=2.033, p<.05
d. t(26)=2.033, p>.05
e. More information is needed to answer the question.
2. A researcher has a sample of 38 people with a mean of 35 and a standard deviation of 7. He thinks the sample
came from a population with a mean of 42. For this study,
a. z=6.16, p<.05
b. z=-6.16, p<.05
c. t(37)=-6.16, p<.05
d. t(37)=6.16, p>.05
e. z=-6.16, p>.05
3. A nutritionist investigates the impact of type of breakfast on mid-morning concentration. He gathers a sample of
50 adults. He feeds them a breakfast and 4 hours later measures their concentration on an interval-level
concentration test. On one day he feeds them bacon and eggs and on a second day he feeds them oatmeal and
fruit. He compares mean concentration on the different types of breakfasts. What type of test should be run in
this study?
a. One-sample z test
b. One-sample t test
c. Paired samples t test
d. Descriptive statistics
e. None of the above
4. A researcher for a health magazine compared the mean caffeine content for a sample of coffees served at coffee
houses to the USDA standard for the mean amount of caffeine in a cup of coffee. What statistical test should he
use?
a. Paired sample t test
b. Related samples t test
c. Dependent samples t test
d. One-sample t test (assuming he isn’t given USDA sigma)
e. z test
5. In Quebec, the average length of time the flu lasts is 6.30 days. An infectious disease physician has developed a
treatment that he believes will treat the flu more quickly. His DV is the number of days the flu lasts. The
hypotheses for the length of time flu lasts should be:
a. Ho: x new treatment ≥ x old treatment ; Ha: x new treatment < x oldtreatment
b. Ho: μnew treatment =μ oldtreatment ; μnew treatment ≠ μ oldtreatment
c. Ho: x new treatment =x old treatment ; Ha: x new treatment ≠ x old treatment
d. Ho: μnew treatment < μold treatment ; Ha: μnew treatment > μold treatment
e. Ho: μnew treatment ≥ μ oldtreatment ; Ha: μnew treatment < μold treatment
f. Ho: x new treatment ≥ μold treatment ; Ha: x new treatment < μold treatment
6. If tcrit=+/-1.734,
a. how many people were in the study, assuming the test is one-tailed and alpha is 0.05?
19 (df=18)
b. and tobs=-1.91, do you reject your null hypothesis? Back your decision up with a numerical statement
(APA style).
Yes! Obs>crit so reject the null, t(18)=-1.91, p<.05
c. If you change your test to a two-tailed test, does your conclusion change? Explain and give the
numerical statement.
Yes. Tcrit changes to -2.101, so you now fail to reject the null. t(18)=-1.91, p>.05.
7. Assume μo=10 and μa=13, with a one-tailed positive end test and alpha=.05. If μa changes to 12, what happens to
(hint: draw a picture!):
a. Power?
b. Type I error?
c. Type II error?
d. Does this make us more or less likely to miss a significant effect of the treatment?
MORE – this is a Type II error.
8. Assume μo=12 and μa=8. Sigma for both distributions is 2. If sigma decreases to 1, what happens to (hint: draw a
picture!):
a. Power?
b. Type I error?
c. Type II error?
d. Does this make us more or less likely to falsely reject the null hypothesis?
It doesn’t affect it – that is a Type I error (alpha), which has not changed in size! But note that we’ve
become more likely to correctly reject a null hypothesis (power has increased).
9.
Based on the table above,
a. What type of test was run? INDEPENDENT SAMPLES T TEST
b. Was the assumption of homogeneity of variance met? YES. LEVENE’S SIG > .05.
c. What is t observed? -1.163
d. Is there a significant difference in height for the two conditions? NO. SIG >.05 FOR T TEST.
e. Does it make sense to calculate effect size? Explain your answer. NO BECAUSE THERE IS NO
STATISTICALLY SIGNIFICANT EFFECT HERE. IF THERE IS NO STATISTICAL EFFECT, WE DON’T TEST FOR A
REAL-WORLD EFFECT (EFFECT SIZE).
f. If you were not given your p value, you could still have drawn the conclusion in d. Explain and show how
you would do this. LOOK UP IN T TABLE FOR DF=11. DEMONSTRATE THAT IT IS GREATER THAN T
OBSERVED.
10. An elementary school principal has just learned of a new method of teaching arithmetic to Grade 2 students.
She is curious to see if the new method is better than the old one. She selects 30 students at random and assigns
them to work with the Grade 2 math teacher. Based on the students’ standardized test scores, the principal
matches them in terms of intellectual ability, forming 15 pairs. The teacher teaches arithmetic by the traditional
method to one member of each pair and uses the new method with the other 15 students. Both groups receive
equal amounts of instruction. Which member of each pair receives which instruction method is determined at
random. At the end of the year, the students take a standardized math test.
a. What test should you use to analyse these data? PAIRED SAMPLES T TEST
b. You calculate tobs=4.78. What does this indicate? Justify your answer. REJECT THE NULL. THE NEW
PROGRAM IS BETTER THAN THE OLD ONE. T-CRITICAL IS 1.761 FOR A ONE-TAILED TEST WITH DF=14.
c. Imagine that power for this test was 0.50. Is this a concern for this conclusion? NOT REALLY BECAUSE
YOU’VE REJECTED YOUR NULL. LOW POWER IS ONLY A CONCERN WHEN YOU’RE WORRIED THAT
YOU’VE FAILED TO REJECT THE NULL WHEN YOU SHOULD HAVE REJECTED.
11. A new brand of automobile tire is advertised to outlast the leading brand, which has a mean lifetime of 80,000
km. A consumer agency randomly selects 6 tires from the manufacturers first production run and tests the tires
on a machine that guarantees even tread wear.
a. What test should you use to analyse these data? ONE SAMPLE T TEST. YOU DON’T KNOW SIGMA.
b. What is tcrit for this test? 2.015, ASSUMING THIS IS RUN AS A ONE-TAILED TEST.
12. In a one-tailed test with alpha=.01 we have rejected the null hypothesis. State whether each of the following is
true or false:
a. The probability that the null hypothesis is false is 0.99.
i. False. We never test how likely it is that the null is false – we test how likely we would be to see
data like this if our null were true. In this case, we’re saying we would see data like this less than
1% of the time.
b. The probability that we made the wrong decision on the null is 0.01.
i. True. Since we’ve rejected the null, we’re looking at our error possibly being a false positive, or a
Type I error, which has a probability of alpha.
c. If we always use the .01 level of significance in our hypothesis testing, over the long run we will make a
wrong decision on the null hypothesis 1% of the time.
i. False. We would have a higher than 1% chance of error, because sometimes we’re going to have
false negatives as well. We need to add them to our chance of a Type I error over time (which
would be 1%).
d. If we always use the .01 level of significance, over the long run we will make the wrong decision on the
null hypothesis on 1% of the occasions when we reject the null.
i. True. See c above.
e. If we always use the .01 level of significance, over the long run we will make a wrong decision on the null
hypothesis on 1% of those occasions when we retain the null.
i. False. If we’ve retained the null, our observed value is in the non-rejection region, which
comprises 99% of our data in this case. Our chance of making an error when we retain the null
comes from Type II error (beta), which is given in the alternative distribution. We don’t have the
information to conclude what it would be in this case, so it could theoretically be 1% but it’s
likely some different value (unless power=. 99).
f. If we always use the .01 level of significance, over the long run we will make a wrong decision on the null
hypothesis on 1% of the occasions when the null is true.
i. True. When alpha is 1%, we’ll accidentally reject a true null 1% of the time. This is the chance of
a Type I error (p=alpha).
13. Dr. Brown is a research director for a school district comprising 2500 students. She reads a test manual stating
that the national norm for Grade 6 students on a math achievement test is 85 with a standard deviation of 10.
She wants to know whether students in her district are performing at the same level. She selects a random 100
students from all Grade 6 students in the district and gives them the same test. They obtain a mean score of
87.1. Follow all the steps of hypothesis testing to test Dr. Brown’s hypothesis.
1. State hypotheses.
Ho: mudistrict=munational Ha: mudistrict ≠ munational
Two-tailed because she just wants to know if they’re the same or different. Emphasize that the data
(which show a mean higher than national data) does not impact the hypotheses that we state.
2. Calculate observed statistic.
We can run a z test because we know population mean and SD.
z=(87.1-85)/(10/root(100)) = 2.1/1=2.1
3. Look up critical value.
Two-tailed test, assume alpha is .05. Need a critical value that cuts off 2.5% of data in each tail.
Zcrit=+/- 1.96
4. Compare observed and critical to make a decision.
Zobs>zcrit so reject the null hypothesis.
5. State conclusion.
The students in her district are significantly better than the national norm, p<.05.
14. Assume that Dr. Brown wants to be more stringent and adjusts her alpha to 0.01 instead.
a. Does this change her conclusion?
i. Yes. Zcrit is now +/-2.58. Zobs<zcrit, so she does not reject the null hypothesis. She concludes
that students in her district are not significantly different from the national norm. Draw a picture
on the board showing zcrit for each alpha and how zobs falls between them.
b. A parent subsequently complains to Dr. Brown that the students are probably smarter but Dr. Brown’s
test wasn’t sensitive enough to see that.
i. In terms of Type I and Type II errors, what is the parent suggesting?
1. Suggesting that Dr. Brown made a Type II error – false negative – she missed a
difference that was really there.
ii. Dr. Brown replies saying that although it was a strict test (alpha=.01), it was powerful
(power=0.98). Should this reassure the parent? Explain.
1. Yes. This means that if the kids in the district were smarter (if the alternative hypothesis
is true, as the parent seems to suspect), there is a 98% chance that the test would have
said so. There is only a 2% chance that they are indeed smarter and that Dr. Brown’s test
missed the difference.
