4 - Categorical data

Question 1

What methods can be used to analyse categorical data?

Answer 1

• Chi2 test of association (Pearson’s)
• Chi2 test of association with continuity correction
• Chi2 test for trend
• Proportions with 95% CI

Question 2

What is the main pitfall of analysing data categorically?

Answer 2

Categories reduces the power of a study and may mask important features of the study by compiling results.

Question 3

What is the purpose of Chi tests of association?

Answer 3

To investigate if two categorical variables are independent of one another, or if there is an association between them by measuring the difference between two proportions (obs and exp) and seeing if it is significant or not.

BUT the extent of any associations cannot be determined.

Question 4

What are the assumptions of a Chi2 test of association, and what should be done if they are not met?

Answer 4

1) observations are independent
2) EXPECTED values are not too small (none <1 and no more than 20% <5)

Use Fishers exact test (if observed values are very small)

Question 5

What is a Chi2 test for trend?

Answer 5

Used when proportions are being compared between two groups and the results are ORDERED! (this method should always be used for ordered data as it can give a statistically significant result when the Chi2 test of association may not)