Categorical data 1

Question 1

When carrying out tests for proportions what makes them valid in a one sample test?

Answer 1

when r>5 and (n-r)>5 where r is the number in the sample with the characteristic of interest

Question 2

In a two-sample test of proportion what is necessary to make it valid?

Answer 2

r1, R2, (n1-r1) and (n2-r2) are all >5

Question 3

What is the simplest approach of carrying out a test of association?

Answer 3

Chi-squared test

Question 4

What is the Chi-squared test?

Answer 4

X2 = sigma (obs-exp)2/exp

Question 5

What are the assumptions for a Chi-squared test?

Answer 5

observations are from independent subjects or experimental units
The expected values are not too small as a rough rule, no more than 20% of the expected values should be <5 and none should be <1

Question 6

What test should be used instead of a Chi-squared test if the variables are ordered?

Answer 6

Chi square test for trend (trend needs to be linear to be appropriate)

Question 7

In the case of a 2X2 table, what test should be carried out when the values of interest are very small?

Answer 7

Chi squared test with continuity correction (conservative) or Fisher’s exact test

Question 8

What tests can be used for comparisons of proportions from paired samples?

Answer 8

Paired confidence interval

McNemar’s test

Question 9

What test should be used for 1 variable with 2 categories?

Answer 9

1 proportion CI and test

Question 10

What test should be used in a 2x2 table?

Answer 10

2 proportion CI and test
Chi-squared test of association
Chi-squared test of association with continuity correction
Fishers exact test

Question 11

What test should be used in >2x>2 table

Answer 11

Chi-squared test of association
Chi-squared test for trend( if categories ordered)
or collapse rows and/or columns to 2x2

Question 12

What test should be used for paired data?

Answer 12

McNemars or CI for paired data