Residuals and Non-normality
This page will give some tips on aspects of looking at residuals in order to assess non-normality. In particular it will stress how
- in repeated measures data, the thing that matters is the distributions of differences between pairs of measures (or more general contrasts in the repeated measures).
- in grouped data, it is the residuals adjusted for group means that matter
Consider this very small example:
X1 |
X2 |
1 |
2 |
2 |
1 |
3 |
2 |
1001 |
1002 |
The distributions of the variables are very skewed. The distribution of the differences (for a repeated measures t-test) are at least symmetrical, if not normal. [Last updated on 21 September, 2010]
Return to Statistics main page
These pages are maintained by Ian Nimmo-Smith and Peter Watson