Lecture 4

Question 1

What are the three types of t tests?

Answer 1

1. Pooled (equal) variance
2. Unequal variance
3. Paired

Question 2

What is the biggest “driver” of the t test?

Answer 2

Sample size (it’s in the denominator)

Question 3

What steps do you take to compare two means?

Answer 3

1. Test each group for normality

2. Perform a t test (choosing equal, unequal, or paired as appropriate)

Question 4

How should you test for normality?

Answer 4

1. Visually: box plot, histogram, q-q plot
2. Numerically: Shapiro-Wilks test or skewness/kurtosis test
3. Assess how close the mean and median are to one another

Question 5

Are t tests robust?

Answer 5

Yes!

Question 6

What tests can be misleading for testing for normality?

Answer 6

Box plots and histograms when your sample size is low.

Question 7

What should you be especially wary of when examining q-q plots for normality?

Answer 7

Stacked points

Question 8

What normality test is best for small samples?

Answer 8

Shapiro-Wilks

Question 9

How is the null hypothesis portrayed in stata?

Answer 9

W0

Question 10

What does the 95% CI give you an idea of?

Answer 10

The spread of values around the mean

Question 11

What can you do if you discover that your data is not normal?

Answer 11

1. Transform the data

2. Use a nonparametric test

Question 12

What transformations would you use for negative skew?

Answer 12

1. X^2 (mild)

2. X^3 (extreme)

Question 13

What transformations would you use for a positive skew?

Answer 13

1. sq rt X
2. ln(X)
3. -1/sq rt X
4. -1/X