The DHS Program User Forum: General » Mismatch in denominator STATcompiler vs analysis of data

Home » Topics » General » Mismatch in denominator STATcompiler vs analysis of data

Show: Today's Messages :: Show Polls :: Message Navigator

Mismatch in denominator STATcompiler vs analysis of data [message #26048]

Wed, 01 February 2023 13:27

Blackandwhite
Messages: 2
Registered: February 2023

Member

Hi,

I am extracting the number of medical injections (variables v477 and v478) for all countries. I like to compare my results to those reported on STATcompiler to make sure I do not make errors. For the vast majority of countries, I get the exact same results. For a few, there is a mismatch in the number of women I have and the number of women on STATcompiler (ie, mismatch in the denominator). Eg: for BJ2017DHS, the number of women listed on STATcompiler is 7712 (unweighted) whereas by analysing the data, I have 15928. Quite a difference. I checked and it does not seem to be due to age, the date of the interview, residence or marriage status. Other suggestions that could explain this? This is not the only country for which I see a difference.
Many thanks for the advice

[Updated on: Wed, 01 February 2023 13:28]

Report message to a moderator

Re: Mismatch in denominator STATcompiler vs analysis of data [message #26099 is a reply to message #26048]

Tue, 07 February 2023 16:19

Janet-DHS
Messages: 893
Registered: April 2022

Senior Member

Following is a response from DHS staff member, Tom Pullum:

I have checked BJIR71FL.dta, and agree with you that n=15928. The mean number of injections and the percentage of women who received an injection (in the last 12 months) in the data file agree with the numbers coming from STATcompiler. However, in STATcompiler I don't see how to get the denominator for this indicator. How do you get the denominator 7712?

Report message to a moderator

Re: Mismatch in denominator STATcompiler vs analysis of data [message #26101 is a reply to message #26099]

Tue, 07 February 2023 16:38

Blackandwhite
Messages: 2
Registered: February 2023

Member

Thanks. It is true that the number of injections match and in this example, I was wondering about the denominator only.

To see it on STATcompiler, you need to select Options in the right-hand corner and then select denominator. I attached a screenshot here (see Benin2017-18).

Other examples where there is a mismatch in the denominator are Haiti 2016-17 (I have 15513; STATcompiler reports 9628), Pakistan 2017-18 (I have 15068; STATcompiler reports 12364) and Timor-Leste 2016 (I have 12607; STATcompiler reports 4348). All these are data for women only.

For other surveys, there is a mismatch in other variables. Eg: % of medical injections done with new/unopened injection material. For Cote d'Ivoire 2005AIS, I get 96.2% and STATcompiler reports 92.1%

Attachment: BENIN DHS.png
(Size: 514.44KB, Downloaded 132 times)

[Updated on: Tue, 07 February 2023 16:39]

Report message to a moderator

Re: Mismatch in denominator STATcompiler vs analysis of data [message #26130 is a reply to message #26101]

Fri, 10 February 2023 08:51

Janet-DHS
Messages: 893
Registered: April 2022

Senior Member

Following is a response from DHS staff member, Tom Pullum:

Thanks for the clarification. Yes, there is a discrepancy. We will look into it and post an explanation--or make a correction to STATcompiler--as soon as possible.

Report message to a moderator

Previous Topic:	what are 707 districts of india covered under NFHS-5?
Next Topic:	Qualitative studies on maternal and newborn health question validity

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sun Dec 1 18:47:04 Coordinated Universal Time 2024