Home » Data » Weighting data » Clustered Standard Errors
Re: Clustered Standard Errors [message #22185 is a reply to message #22173] |
Tue, 09 February 2021 12:24 ![Go to previous message Go to previous message](/theme/default/images/up.png) ![Go to next message Go to previous message](/theme/default/images/down.png) |
analyst_till
Messages: 10 Registered: January 2021
|
Member |
|
|
Thanks for the tips concerning the Strata.
With respect to the districts: Although my unit of analysis is the individual, my identification strategy relies on changes at the district-level. Therefore, I include district fixed effects in my regression framework and as far as I know, these are reflected by "i.district".
Please correct me if I am wrong, but I guess this term is crucial to guarantee that I compare only observations of the same districts over time?
As far as I know, STATA breaks the arising collinearity between i.district and i.treatment by dropping one of the district fixed effects.
Also, I am still a bit confused about the clustered standard errors. According to Duflo et al. 2004, one can tackle the issue of serial correlation in a diff-in-diff by clustering standard errors at group-level, allowing for autocorrelation within the groups.
Since the group/cohort in my panel is district, shouldn't I cluster standard errors at district level?
If standard errors are clustered at PSU-level through "svyset", I would not cluster SE at the level of my cohort.
Is it possible that I have to code?:
svyset districts [pweight=v005], strata(stratum_ID) singleunit(centered)
Sorry if it was not clear, I hope you understand what I mean!
Best,
Till
[Updated on: Tue, 09 February 2021 12:25] Report message to a moderator
|
|
|
Goto Forum:
Current Time: Sat Aug 31 05:24:05 Coordinated Universal Time 2024
|