I'm trying to estimate the effect of an event. The data I have is number of sales in 7 different periods of time (3 periods before the event, the period that the event happens and 3 periods after the event). I have a relatively small group to that participate the event and a much bigger that didn't with similar characteristics. When we performing a difference in difference we generate a control group dummy $treat_{i}$ which is equal to 1 if a person is in group A and 0 otherwise, then we generate a time dummy $time_{t}$ which is equal to 1 if t=2 and 0 otherwise, and then we regress:

$Y_{it} = \beta_{1} + \beta_{2}(treat_{i}) + \beta_{3}(time_{t}) + \rho (treat_{i} \cdot time_{t}) + \epsilon_{it}$

So my questions are:

  1. Do we assume $Y_{it}$ have a normal distribution?
  2. Should I use a grouped variable such as mean or median for $Y_{it}$ over time or the whole data?

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.