4 Choosing the correct test

Question 1

What does a t-test do?

Answer 1

Compares means between 2 independent groups

Question 2

The confidence interval is as follows: (-20.8, 60.7). What does this tell us about the null hypothesis?

Answer 2

95% CI crosses 0 = null hypothesis

Question 3

What 4 questions do we need to ask to decide what type of test to use?

Answer 3

1. Outcome variable?
2. Normally distributed?
3. Are observations correlated?
4. How many groups are being compared?

Question 4

In what case would you use a t test?

Answer 4

Continuous outcome variable
Normal
Independent observations
2 groups compared

Question 5

In what case would you use a paired t test?

Answer 5

Continuous outcome
Normal
Correlated
2 groups

Question 6

What does a paired t test do?

Answer 6

Compares means between 2 related groups

Question 7

What is the difference between t test and paired t test?

Answer 7

T test used when groups are unrelated, paired t test when groups are related

Question 8

If you have a continuous variable, normally distributed, obs are unrelated and you are comparing 3 groups, what test are you using?

Answer 8

ANOVA

Question 9

What does ANOVA stand for?

Answer 9

ANalysis Of VAriance

Question 10

What does a p<0.05 in ANOVA test tell you?

Answer 10

It doesn’t mean all the groups are statistically different, it just means one of them is

Question 11

What can we use to correct for multiple comparisons?

Answer 11

Bonferroni correction

Question 12

When would you use the Wilkoxon sum rank test?

Answer 12

Not normally distributed alternative to t test

Question 13

When would you use Kruskal Wallis test?

Answer 13

Not normally distributed alternative to ANOVA

Question 14

What does the chi squared test do?

Answer 14

It compares proportions between 2 groups or MORE

Question 15

In what cases would you use Fisher’s exact test and McNemar’s exact test?

Answer 15

Cells < 5
If independent groups = Fisher’s
If related = McNemar’s

Question 16

When would you use McNemar’s chi square test?

Answer 16

When the observation is correlated and not sparse cells

Question 17

If a study looks at before and after, what test are you going to use?

Answer 17

Paired t test (before and after are correlated)

Question 18

Whats the difference between Wilkoxon SUM and Wilkoxon SIGN rank test?

Answer 18

SUM is for t-test (independent)
SIGN is for paired t test (correlated) (think of the plus sign)

Question 19

What is the problem with Benferroni correction?

Answer 19

It assumes complete independence between comparisons - too consrevative