Differential dropout analysis in longitudinal studies

MaryJane — Mon, 26 Apr 2021 12:34:52 GMT

Hello,
I have a dataset with patients whose measurements were recorded three times during a study. Many participants are completers, they showed up during all measurements (all three time points). Others dropped out earlier. How can I plot the means, not per control and treatment group only but also stratified per time point completion? In other words, I would like to stratify the different group's trend per dropout time. For instance, I would do these in two panels. Each panel or window will show the placebo vs therapy group. On each graph within each panel, one will be able to see the mean response of participants that dropped out at the next stage. Ex. If the graph turns out to show a group mean lower for early study dropouts than later ones, then I can already get an idea if the model I am fitting is likely to produce biased estimates.

For info, I am including a sample screenshot of graphs I would like to produce in the context of my dataset. The attached graphs are just examples I have found on the internet, which I would like to reproduce in SAS.

Thank you in advance.

Re: Differential dropout analysis in longitudinal studies

ballardw — Mon, 26 Apr 2021 16:31:35 GMT

Do you have a variable that indicates the drop out time?

If so sort by your "treatment" and that variable, use Proc SGPanel with the treatment and drop out time variable as Panelby variables.

At least that is where I would start.

topic Differential dropout analysis in longitudinal studies in Statistical Procedures

Differential dropout analysis in longitudinal studies

Re: Differential dropout analysis in longitudinal studies