Difference between attritors and subsequent joiners in longitudinal panel

[Max Brauer]

I am running an attrition analysis of panel data. I am trying to determine if there is a difference across a variety of characteristics between panelists who attrited from a study and subsequently joined the study in later waves. I’m assuming I could just run a ttest to compare the differences between the two groups across the various variables. How would I do this?

I am away from my desk and not at able to post data from dataex at the moment, but can follow up if needed.

[Clyde Schechter]

Well, if you want to compare the groups on baseline measurements of those characteristics, or if those characteristics are immutable, then, assuming you select only one observation per panelist to work with, then, yes, you can use a t-test. But the nature of panel data is that you have repeated observations. If the characteristics you want to compare can change over time, and you want to look at more than just the baseline measurement, then you will have to include multiple observations per panelist, and the t-test will not be valid.

If you want more specific advice, you will need to show example data, and describe specifically the comparisons you want to make.