Home » Topics » Mortality » accounting clustering effects of women's data when using baby-based analysis
| Re: accounting clustering effects of women's data when using baby-based analysis [message #794 is a reply to message #793] |
Sat, 28 September 2013 05:43   |
rkinoshita
Messages: 6
Registered: September 2013
Location: Nicaragua
|
Member |
|
|
Hi. Thank you so much for your prompt response. This is indeed very helpful. I really appreciate it.
sorry for being insistent, and asking more questions but I do have a few follow up questions for you.
1) I run the cluster command that you mentioned in your point 1, and it does not work for some reason. I past my STATA output here. IPVlife- my exposure variable, Infant- infant deaths for all births. HHCLUST= PSU. I thought that it did not work because this command is only for regression rather than logistic, but it still did not work. any thoughts would be great.
. xi: svy: logistic IPVlife infant [pweight=weight], cluster(HHCLUST)
weights not allowed
r(101);
. xi: svy: logistic IPVlife infant [pweight=weight2], cluster(HHCLUST)
weights not allowed
r(101);
. help cluster
. xi: svy: logistic IPVlife infant [pw=weight2], cluster(HHCLUST)
weights not allowed
r(101);
. xi: svy: regres IPVlife infant [pweight=weight2], cluster(HHCLUST)
weights not allowed
r(101);
2) I do not fully understand what is the difference between -- a random effects method and a "clustering" correction (non-parametric V/C matrix) that you mentioned in your first suggestion. But if I understand well, you are not recommending vce option but using reg Y X....that you suggested, and the reason for this is that vce will only consider the clustering at individual households but not at the bigger level (e.g. PSU level), correct? Does this mean vce option cannot use for clustering at PSU, or in other words, I cannot use vce option with varname PSU??
3) for your second suggestion, I do not understand the weight in "women recode" that you mentioned. in my merged dataset, I have pesomef and pesonino - two weighting variables- one for women and children. Are you talking about these variables? how do you compare these variables (obviously I cannot list it) and how do you use adjust the weight? Do I create another weight variable using pesonino in the merged dataset? FYI, before merging the two datasets (women and baby), each dataset was already accounted for weight.
below is what I did for women, and the same thing in baby's dataset using pesoNino
gen weight2= PesoMEF
svyset HHCLUST [pw=weight2], strata(HHDEPAR)
thanks very much again for your help.
Rinko
|
|
|
|
Current Time: Wed Jan 1 13:31:41 Coordinated Universal Time 2025
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
