Adjusting sample weight for multilevel analysis- SPA dataset [message #15381] |
Mon, 09 July 2018 03:03 ![Go to next message Go to next message](/theme/default/images/down.png) |
rkchettri
Messages: 19 Registered: November 2017
|
Member |
|
|
Dear DHS and STATA experts,
I am using Nepal Health Facility Survey data to identify the determinants of quality of care of ANC service. My unit of analysis will be the pregnant women who visited health facility for ANC services.
While checking datasets I found that more than one pregnant women were sought ANC services from each health facility, and health facility and ANC users have different sample weight. In this situation, I how can I adjust sample weight for those ANC users and health facility. I am planning to analysis multi-level linear regression.
With kind regards,
Resham
|
|
|
Re: Adjusting sample weight for multilevel analysis- SPA dataset [message #15764 is a reply to message #15381] |
Tue, 11 September 2018 11:15 ![Go to previous message Go to previous message](/theme/default/images/up.png) |
Bridgette-DHS
Messages: 3139 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Senior Research Associate, Wenjuan Wang:
The simple answer is that we don't have separate sampling weights for each level, in this case, for the selection of facilities within a stratum and the selection of clients within a facility. There is a similar issue when people try to run multilevel models with DHS data. SPA has a client weight; but its computation takes the multi-stage selection into account, not just the selection of clients in the second stage. So it is not the ideal weight to use for the second level. In past work, we have used facility weight as the first level weight, and assumed equal weight for all clients in the same facility (which may not be true). In your case, you would let clwt=1. This is the best we can do and other researchers may have done it differently.
|
|
|