Analysis using a categorical mixed model

The number of post-treatment epilepsy attacks was categorised into four groups as shown below. The groupings are chosen so that each category is of a reasonable size.

Group

Attacks

1

0 (10%)

2

1-3 (29%)

3

4-10(39%)

4

11+ (22%)

The baseline attack rate was not categorised and was again fitted as log(attacks). Attempts were made to fit compound symmetry, Toeplitz and general covariance patterns to the data using the SAS macro written by Lipsitz etal. (1994) (see Section 9.3). However, convergence was achieved only for the compound symmetry model. The treatment-time interaction was not significant and this effect was removed from the model.

Was this article helpful?

0 0

Post a comment