Here are the answers to the exam questions. The answer is in **bold** and the explanation for the answer is in regular type. If you have quick questions about your exam, you can talk with me before or after class. If you'd like to talk with me about the exam outside of class, e-mail me to set up a time to meet.

**1. ** One nominal variable, mouse strain, with more than two values; one measurement variable, exercise time: ** Fisher's one-way anova ** or ** Welch's one-way anova **

**2. ** Two nominal variables, quadrat that the butterfly flies to; magnetic field blocked or not; total sample size less than 1000; ** Fisher's exact test of independence **

**3. **One nominal variables, male vs. female; one measurement variable, neck length; experiment is unbalanced (28 females and 52 males) and heteroscedastic (standard deviations are 0.41 and 0.23); **Welch's one-way anova** or **Welch's t-test**.

**4. ** Two nominal variables, habitat, male vs. female; total sample size less than 1000; ** Fisher's exact test of independence **

**5. ** One nominal variable, sleeping position, with more than two values; one measurement variable, snoring time: ** Fisher's one-way anova ** or ** Welch's one-way anova **

**6. ** One nominal variable, what happens to the turtle; one ranked variable, the order the turtles emerge: ** Kruskal-Wallils test. **

**7. ** Two nominal variables, species of isopod, pine vs. oak forest:

** Fisher's exact test of independence
chi-square test of independence
G-test of independence **

**8. ** One measurement variable, distance moved; theoretical expectation (mean distance is 0) if the null is true; total sample size is 62 turtles; ** Student's one-sample t-test. **

**9. ** One nominal variable, treatment; one measurement variable, plant height. So you would start with a one-way anova. But to just compare one pair of means, out of the 15 possible pairs, you would follow up the Fisher's one-way anova with the ** Tukey-Kramer test. ** You could have said you'd follow up the Welch's anova with the Gaines-Howell test, but you're not required to know Gaines-Howell for this class.

**10. ** One nominal variable, kind of corn; theoretical expectation (1/3 of corn borers in each kind of corn); total sample size 120: ** exact test of goodness of fit **

**11. ** Three nominal variables, one vs. two legs; morning vs. evening; day of the week: ** Cochran-Mantel-Haenszel test **

**12. ** One nominal variable, the kind of male that each female mates with; theoretical null expectation; total sample size 1400: ** chi-square test of goodness-of-fit ** or ** G test of goodness-of-fit **.

**13. ** ** alpha: use 0.05 because most people do
power: use 0.90 because that's a nice round number that a lot of people use (or beta=0.010)
effect size: pick a number based on the change in collagen content that I would think is interesting
standard deviation of collagen content: either use a number from the literature on other rabbit collagen experiments, or do a pilot study to measure this **

**14. ** **In 50 experiments testing null hypotheses that are all really true, you will expect about 2.5 experiments to give you significant (P<0.05) results. So you can pick out a couple of experiments where the no-touch reiki "works" (gives a P<0.05) and use those results to convince gullible people that it works. ** (And "no-touch reiki" is a real thing; some people even claim they can channel the healing life forces of the universe via an expensive *phone call*.)

**15. ** ** If the error bars are standard deviations, they tell you that there is more variation among individual tail lengths in DBA/2 mice than the other mice.
If the error bars are standard errors, they tell you that the estimate of the mean tail length for DBA/2 isn't as good as the estimate for the other mice.
If the paper doesn't say what the error bars represent, it tells you that the authors of the paper are either careless, or don't know very much about statistics.**