6
$\begingroup$

I would like to analyse data with a crossover design. Each subject was assigned to a sequence (AB or BA), i.e. 2 periods. Each period consisted of one measurement before treatment (baseline or pre-treatment) and one after treatment.

The data look like:

subject | sequence | period | treatment | time | result  
   1         AB        1         A        pre      5  
   1         AB        1         A        post     7  
   1         AB        2         B        pre      6  
   1         AB        2         B        post     5  
...

I would like to compare pre- and post-treatment results for each treatment. My attempt was to use a mixed model with treatment, sequence, period, time (pre or post-treatment) and interaction between time and treatment.

In R, this would look like:

lme(result ~ treatment + sequence + period + time + time*treatment, random = ~1|subject)

Is this a correct way for comparing the pre- and post-treatment results in a crossover design?

$\endgroup$

0

Browse other questions tagged or ask your own question.