Dear all,

I've collected survey data within a multi-armed RCT format, which consists of one control frame and three different treatments. Before and after exposure to the treatment respondents were asked the same set of questions, so I'm looking to see whether the treatment they received affected their answers. I am assuming a DID estimator would be appropriate to use here.

Is it correct to include each of the treatments variables in the regression as well as their interaction variables with time, or would I need to regress each treatment separately?

For the former: reg Outcome Time Dummy1 Dummy2 Dummy3 Time*Dummy1 Time*Dummy2 Time*Dummy3 ControlVariables
For the latter: reg Outcome Time Dummy1 Time*Dummy1 ControlVariables

Where 'Dummy#' is the dummy variable indicating which treatment they received, and 'Time*Dummy#' is the interaction between time and the treatment they received.

Furthermore, in general, when comparing results from a multi-armed RCT such as this what is the best way to deal with the potential for type I error increasing with multiple hypothesis testing - A Bonferroni correction, or something else? If so, how would could I account for this within STATA?

If anything is unclear I'd be happy to elaborate and explain further.

Many thanks in advance for your help.

Ruby