Hypothesis Tests: Categories

Question 1

summarize which test to use, given the predictor data type and outcome data type

Answer 1

Question 2

describe chi-squared tests

Answer 2

• chi-squared tests are for categorical outcomes and predictors, when you have no information on variance on the predictor (chi-tegorical)
• the null hypothesis is that the predictor varies randomly by category of the outcome
• the alternative hypothesis is that the predictor differs nonrandomly by category of the outcome

Question 3

describe the equation for Chi-squared

Answer 3

Question 4

describe T-tests

Answer 4

• T-tests are for comparing the means of two groups
  
  • (Tea is for Two)
• the null hypothesis is that the 2 groups have the same mean value of the predictor
• the alternative hypothesis is that the 2 groups have different mean values of the predictor

Question 5

explain how a T-test works

Answer 5

a T-test answers the question, “how many standard error apart are these 2 sample means?”

• if many, the true means are probably different; reject the null hypothesis
• if not many, the true means may be the same; accept the null hypothesis

the p-value tells, “how likely is it that these 2 sample means would be this far apart, if they are in fact drawn from the same population?”

Question 6

describe details and assumptions of T-tests

Answer 6

• T-tests compare population means (not any other measure of central tendency); they assume the sample population is normally distributed

Question 7

describe ANOVA

Answer 7

T-test for 3 or more groups

• ANOVA partitions the total variance into within-group and between-group
• the null hypothesis: the means of all 3 categories are not significantly different (that all variance is within-group)
• the alternative hypothesis: at least 1 category is different (that at least some variance is between-groups)

Question 8

name 1 way to decrease within-group variation

Answer 8

one way to decrease within-group variation is with a paired test

• paired tests measure the same subjects with and without the predictor of interest
  
  • this decreases within-group variation, since all within-subjects traits are corrected for

Question 9

list the 3 ways to increase statistical power

Answer 9

1. increase sample size
2. decrease within-group variance
3. sample only the most extreme individuals (the “best examples” of each group)

Question 10

summarize the types of tests, in increasing order of power

Answer 10