The DHS Program User Forum: India » Appropriate stratification variable for NFHS data

Home » Countries » India » Appropriate stratification variable for NFHS data

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

Appropriate stratification variable for NFHS data [message #8180]

Fri, 04 September 2015 14:17

econ_stata
Messages: 5
Registered: July 2015
Location: India

Member

I am working with the NFHS-3 survey KR dataset in Stata, studying the determinants of child health at the all-India level, and have a doubt about the appropriate stratification variable for the survey declaration. Looking at various answers on the Forum, I see recommendations by experts for both

1. v022, "sample stratum number" (73 distinct values)

e.g, by Trevor-DHS here

http://userforum.dhsprogram.com/index.php?t=msg&goto=326 8&S=c572d650be37c4375bff4d100ab9ced0&srch=NFHS+strat a#msg_3268

2. a grouping of v024, "state" (29 values) with v025, "type of place of residence" (2 values), i.e.,

egen float stratum = group(v024 v025), resulting in 58 distinct values,

e.g., from Ruilin Ren here

http://userforum.dhsprogram.com/index.php?t=msg&goto=213 4&S=c572d650be37c4375bff4d100ab9ced0&srch=NFHS+strat a#msg_2134

I was using option 2 so far, but revisiting this issue, I checked the NFHS publication

'National Family Health Survey (NFHS-3), 2005-06: India: Volume I', section on Sample Design, p61, (copy uploaded)

and it seems the stratification scheme is more complex than just the simple grouping of State & place of residence.

I would really appreciate a definitive answer to this from one of the experts familiar with the NFHS data- I think this would serve many researchers now & in future as well.

Attachment: 2SamplingDesign_from_NFHS_FRIND3-Vol1AndVol2 copy.pdf
(Size: 203.07KB, Downloaded 555 times)

Thank you,
Regards

[Updated on: Fri, 04 September 2015 14:24]

Report message to a moderator

Re: Appropriate stratification variable for NFHS data [message #8232 is a reply to message #8180]

Fri, 18 September 2015 14:20

Bridgette-DHS
Messages: 3199
Registered: February 2013

Senior Member

See below the answer from Senior DHS Sampling Specialist, Ruilin Ren.

The exact stratification of the NFHS-III was very complicated, it is state specific. We do not have the complete information for the exact stratification. I am not aware of the variable HV022 with 73 codes, it might be inserted to the data file much later (on 2010). I will check with other colleagues to figure out what HV022 stands for in NFH-III.

Since the exact stratification is not available, as usual, we always suggest to use HV024 (state) cross HV025 (urban/rural) as approximate stratification. Since there are 8 cities in India which have slum areas, the slum areas were oversampled in NFHS-III. If you want and if your analysis is sensitive to slum and non-slum areas, you can further split the HV024 by HV025 stratification by adding SHSLUMC (variable indentifies slum areas defined by census) at HH level, or using V024 by V025 by SSLUMC at individual level. The eight slum cities are as follows:

- Delhi in the state of Delhi
- Kolkata in the state of West Bangal
- Hyderabad in the state of Andhra Pradesh
- Chennai in the state of Tamil Nadu
- Indore in the state of Madhya Pradesh
- Merrut in the state of Uttar Pradesh
- Mumbai in the state of Maharashtra
- Nagpur in the state of Maharashtra

Report message to a moderator