Home » Data » Weighting data » Clustered Standard Errors
Re: Clustered Standard Errors [message #22277 is a reply to message #22269] |
Tue, 23 February 2021 07:08 ![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) |
Bridgette-DHS
Messages: 3151 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
For the SMGL analysis in Uganda, and for an earlier analysis in Rwanda, the svyset adjustments were for PSUs even though the units for the difference-in-differences model were districts. I really don't think there is a problem with this approach. Say that you had any binary predictor, such as urban/rural, and two surveys, and you wanted to test whether the change (in some outcome) between the two surveys was the same in both urban and rural areas. You would be testing for the significance of the interaction term. Or maybe your predictor was not spatial at all, for example a binary version of education (e.g. <=primary and > primary). You would test the significance of the interaction term. Isn't your situation equivalent to that? Your binary predictor is whether the person/household is in or is not in an intervention area, and you want to test whether the change between the two surveys was the same in both the intervention areas and the control areas.
The svy adjustments are intended to compensate for how a DHS sample deviates from a simple random sample. If the district ID is not relevant to the sampling design, I don't see why you need to include district ID in the svyset command.
|
|
|
Goto Forum:
Current Time: Sat Aug 31 01:27:07 Coordinated Universal Time 2024
|