The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Merging data files » Mismatch between reported and generated output on merged file
Mismatch between reported and generated output on merged file [message #18177] Sat, 05 October 2019 05:58 Go to next message
marian is currently offline  marian
Messages: 11
Registered: August 2019
Member
Hi,

I have merged two recode files--IR with PR and wanted to check nutritional status across mothers' low BMI (BMI<18.5) but I am not getting the same results as reported in state reports of NFHS 4. I have used the same filters that would have been used for each variable in their respective individual recode files (PR, IR) for NFHS 4.

Eg for Assam, I am getting stunted children of underweight mothers as 42.8% instead of 43.1 and underweight children as 39.9% instead of 40.1%. Also when I had merged the two datasets, only 30% of observations merged. The weighted sample size of state when BMI (tab v445) was checked on this dataset is 32639

Please suggest where the problem lies. I have attached the code.

  • Attachment: IR Merge.txt
    (Size: 1.04KB, Downloaded 50 times)

[Updated on: Sat, 05 October 2019 06:04]

Report message to a moderator

Re: Mismatch between reported and generated output on merged file [message #18243 is a reply to message #18177] Fri, 18 October 2019 14:43 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 1770
Registered: February 2013
Senior Member

Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

The anthropometry variables in the PR file refer to all children in the household. Some of those children do not have a living mother in the same household. The children in the IR file will only be those whose mother is alive and in the same household. This is a slightly smaller number than the children in the PR file.

If this is not the problem, another possibility is that some mothers in the PR file did not get into the IR file. In every survey, there are some women who were identified in the household survey as being eligible for the women's survey, but in fact they were not interviewed, perhaps because they refused (or for another reason).

Another possibility relates to de jure versus de facto residence, given by hv102=1 and hv103=1, respectively. All women and children are measured, without respect to whether they are later classified as de jure or de facto. However, eligibility for the women's interview is limited to women who are de facto residents. (That's the usual DHS criterion; for MICS it is de jure residence; and it's possible for a specific survey to deviate from the norm.) The tables in the main report are mostly limited to de facto residents.


Re: Mismatch between reported and generated output on merged file [message #18264 is a reply to message #18243] Tue, 22 October 2019 05:36 Go to previous messageGo to next message
marian is currently offline  marian
Messages: 11
Registered: August 2019
Member
Thank you Mam..

As per your suggestion, I thought I would look at KR file itself but i checked across and there are differences in sample and output based on filters (flagged cases, pregnant mothers exclusion, exclusion of mothers who had a birth in 2 months prior to survey as NFHS 4 report mentions)..

I have put up tabulation differences (both in sample size and output) for 4 states and its perplexing to find varying differences but I am not getting any result (even if it matches NFHS 4 report) that matches the NFHS 4 reported same sample size. (attached is word file).

The same problem in mismatch occurs when i check nutritional status of children across birth order and birth size.. This only happens when I am do such cross tabulations else individual variables have no problem.. I want a syntax that can be used irrespective of state.

Please guide

[Updated on: Wed, 23 October 2019 06:05]

Report message to a moderator

Re: Mismatch between reported and generated output on merged file [message #18331 is a reply to message #18264] Mon, 11 November 2019 09:09 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 1770
Registered: February 2013
Senior Member

Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

In order to try to answer your question I would have to go through more steps than I have time for. Very sorry about that. I suggest that you do the analysis the way you think it should be done and not worry about whether you match with the tables in the report.
Previous Topic: merging Household members to women file
Next Topic: Merge PR and KR in wave 3, b16 does not exist
Goto Forum:
  


Current Time: Fri Apr 10 15:00:41 Eastern Daylight Time 2020