The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Dataset use in Stata » Difficulty analysing IR data for many countries
Difficulty analysing IR data for many countries [message #1351] Mon, 17 February 2014 06:44 Go to next message
Lorretta is currently offline  Lorretta
Messages: 8
Registered: August 2013
Location: Nigeria
Member
I am using individual recode dataset for Swaziland, Gabon, Cote d'Ivoire, Kenya, Liberia, Namibia, Cameroon, Lesotho, Ghana and Mozambique for a study. Except for Cote d'Ivoire, binary logistic regression using xi:svy:logistic ... gives me only the odds ratio, with a note that missing standard errors because of stratum with single sampling unit. The F and prob > F are without any values. I defined the strata and psu for each country as svyset [pw=wt], psu(v021) strata(v022). Does anybody know what I am not getting right? Thanks.


Lorretta
Re: Difficulty analysing IR data for many countries [message #1362 is a reply to message #1351] Mon, 17 February 2014 19:05 Go to previous message
Reduced-For(u)m
Messages: 292
Registered: March 2013
Senior Member


A couple possibilities:

1 - you need to create a new variable with PSU that is separate for each country/survey-round, same for strata. So, if Ghana has PSUs 1-90 and Lesotho has 1-28 (just made up numbers), you need to tell Stata that these are separate PSUs. You could do this by, say, generating a 2-digit number for each country, multiplying it by 1000, and adding on the PSU number (assuming never more than 3 digit PSU numbers - and something similar for strata).

2 - you say you defined it for each country - does that mean you re-did the svyset each time? Are you running the regressions separately? If you are running them separately, ignore (1) and try (3). If not, re-try (1) only once after all of your data is loaded.

3 - are you using some subsample, or trimming your sample extensively? You might try:

bysort strata: tab Y

and see if there are any strata with 1 ob.

Happy to hear back and offer any other thoughts if this doesn't work or get you on the right path. But my guess is it has something to do with your svyset, or you are using a small subsample (on purpose, or somehow accidentally for some particular country/region/urban (strata). Of course, I've been known to be wrong so....

Previous Topic: Pooled Datasets - Use of Svyset & regional controls
Next Topic: I need help on using select to reduce DHS variables
Goto Forum:
  


Current Time: Thu Mar 28 14:25:37 Coordinated Universal Time 2024