I am using two waves of survey data (not a panel) with the outcome defined at the individual level and the main explanatory variable is an enumeration area level variable. I'm not sure whether the right approach is to cluster standard errors by the enumeration area (since this is the "treatment" level) and include a survey wave fixed effect or to cluster at the wave level.
This relates to me perhaps not understanding the correct fixed effects to use. With only one wave of data, I included a country*region fixed effect to identify within region variation. I'm unsure now though whether to include the previous and a wave fixed effect
Code:
absorb(i.contreg i.wave)
Code:
absorb(i.contregwave)
0 Response to Clustering with multiple survey waves
Post a Comment