The DHS Program User Forum - RDF feed
https://userforum.dhsprogram.com/index.php
Sampling weights harmonization between DHS and MICS
https://userforum.dhsprogram.com/index.phpindex.php?t=rview&goto=17946&th=8289#msg_17946
Kind regards,
Hamid Oskorouchi, Ph.D.
University of Göttingen, Germany]]>hamid2019-07-25T09:57:47-00:00Re: Sampling weights harmonization between DHS and MICS
https://userforum.dhsprogram.com/index.phpindex.php?t=rview&goto=18012&th=8289#msg_18012
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
We apologize for the delay with this reply. Perhaps you have already found an answer. The DHS and MICS surveys have very similar two-stage sampling designs and the weights are calculated in the same way, to compensate for over- and under-sampling of the various strata and for variations in nonresponse. Using the weights will produce unbiased estimates.
In general we discourage calculating pooled estimates using multiple DHS surveys, and for the same reason would discourage pooling DHS and MICS surveys. But if you do want to pool, please check previous posts on the DHS forum about possible adjustments. The issues for pooling a DHS survey and a MICS survey would be the same as for pooling two DHS surveys.
]]>Bridgette-DHS2019-08-16T18:19:53-00:00Re: Sampling weights harmonization between DHS and MICS
https://userforum.dhsprogram.com/index.phpindex.php?t=rview&goto=18015&th=8289#msg_18015
thank you very much for your reply. I believe that my original post was not clear enough.
What I meant by "pooling" is appending and using in the same analysis multiple DHS/MICS of a single country collected at different points in time.
I understand that jointly analyzing multiple DHS from different countries would produce(statistically) meaningless results.
Please, let me know if I am still not getting this issue right.
Best,
Hamid ]]>hamid2019-08-16T19:52:35-00:00Re: Sampling weights harmonization between DHS and MICS
https://userforum.dhsprogram.com/index.phpindex.php?t=rview&goto=18061&th=8289#msg_18061
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
Hi Hamid--I often pool (append) surveys in the same way, for data processing convenience. When I do this, I don't change the weights at all. DHS always normalizes the weights in a file, so that the total weight equals the number of cases (with a factor of 1000000). Stata, when using pweight, always normalizes as well (without a factor of 1000000). If you use v005 (or hv005, etc.) and pweight, in an appended file, you should be ok without changing the weights in any way. If you have any doubts about this, and can think of another way to handle the weights that would make sense, I suggest you try the alternative and compare the results (using the given weights and your adjusted weights) to see if there's any difference. You want to compare both point estimates and standard errors (or confidence intervals) using the two alternatives. If you do get a difference, please let me know. Tom]]>Bridgette-DHS2019-09-03T18:43:17-00:00