2
$\begingroup$

I have 54 individuals, 27 in the treatment group and 27 in the control group. I observe everyone 6 times. However, my treatment is applied in an alternating fashion. In other words, my observations for the treated group are untreated state, treated state, untreated state, treated state, untreated state, treated state. Can someone explain how to set up a difference in difference model with this type of data structure?

The only option I've come up with is to average the observations for the individuals in the treated group when they are in their untreated state and act as if that average is a single "pre-treatment" observation (reducing my number of observations from 27 individuals X 3 time periods to 27 X 1). I could then average the observations for the individuals in the treated group when they are in their treated state and act as if that average is a single "post-treatment" observation (reducing my number of observations from 27 individuals X 3 time periods to 27 X 1). I could then create two similar average observations from my control group, averaging every other observation even though there is no alternating treatment.

Does anyone have suggestions on how to handle this data without collapsing everything in time-serial terms? Is using each treated individual as 3 pre-/post- observations legitimate? How would I account for the lack of independence across three observations of the same individual?

Thanks.

Brian

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.