Is there a carry-over effect when using a mixed-effect model?
Example: Trial with two groups: Intervention and controls. The treatment is applied at several time points, and the effect of the treatment is measured at each point.
A mixed-effect model for repeated measures is used.
Between each time point, one must speculate that there could be a carry-over effect of the treatment from the previous time point.
If you're not able to alter the study design, e.g., the time between each time point, how can you test a carry-over effect?
I am using Stata.
I may not be understanding what you want. But there is no implied carry over from the intervention in the fixed effects estimates of the standard model, although there may be a dependence over time in the response or more correctly the residuals ( in its simplest form a compound symmetry covariance structure see http://www.statpower.net/Content/GCM/Lectures/SW07.pdf ).
That said , you may include the contemporaneous intervention and the lagged intervention; you will of course lose observations with this method as there will be no lagged intervention on the first occasion. Power will depend on the number of observations and the degree to which the intervention is approaching bring time invariant.
- Badges
- Science topic