The DHS Program User Forum: Nutrition and Anthropometry

Home » Topics » Nutrition and Anthropometry » sample weight

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

sample weight [message #2144]

Mon, 12 May 2014 06:25

emarini
Messages: 6
Registered: May 2014
Location: Univerity of Cagliari, It...

Member

Hi, I need help in selecting the right sample weight for analyzing anthropometric data on nutritional status in children. Do you know if the 'HV005' variable is appropriate?
Thank you.
Elisabetta

Report message to a moderator

Re: sample weight [message #2149 is a reply to message #2144]

Mon, 12 May 2014 13:24

Reduced-For(u)m
Messages: 292
Registered: March 2013

Senior Member

I usually use V005 while doing my child anthro analysis on the birth recode. Are you using the household recode (hence the HV)? I just tend to find that recode clunkier when my unit of observation is the child.

Report message to a moderator

Re: sample weight [message #2150 is a reply to message #2149]

Mon, 12 May 2014 13:53

emarini
Messages: 6
Registered: May 2014
Location: Univerity of Cagliari, It...

Member

Thanks. I am quite new with these datasets. I am using the Household Member Recode, because only in this one I found all the variables I have to take in account.

Report message to a moderator

Re: sample weight [message #2151 is a reply to message #2150]

Mon, 12 May 2014 13:59

Reduced-For(u)m
Messages: 292
Registered: March 2013

Senior Member

I think that HV005 is fine then. Just FYI (not totally relevant, the household member recode should be fine) - you can always take data from one recode, save it along with the household ID number, and then merge it into another dataset.

Quick thing though - I suggest using the new WHO standards, which are not provided in many of the datasets, but can be easily computed by downloading a package like "zscore06" for Stata and using the raw measurements, age, and gender variables. In the new datasets (post-2008 or so) there is a separate variable with the WHO standards already there.

Report message to a moderator

Re: sample weight [message #2153 is a reply to message #2151]

Mon, 12 May 2014 14:12

emarini
Messages: 6
Registered: May 2014
Location: Univerity of Cagliari, It...

Member

thanks! it is clear.
I am using the new WHO standards, thanks for the suggestion.

Report message to a moderator

Re: sample weight [message #2161 is a reply to message #2153]

Wed, 14 May 2014 07:15

emarini
Messages: 6
Registered: May 2014
Location: Univerity of Cagliari, It...

Member

Hi. Maybe someone can help me with a doubt on the use of sampling weights in MICS datasets. I am studying child malnutrtion and I am using the unit of analysis "Mothers or primary caretakers of children under the age of five (ch.sav)", with the variable weight "chweight: Children's sample weight". Is the correct one? are there particular rules to be applied for weighting? I do not find a detailed procedure description in the MICS documentation, nor I find a forum such as this one.

Report message to a moderator

Re: sample weight [message #2164 is a reply to message #2161]

Wed, 14 May 2014 15:54

Reduced-For(u)m
Messages: 292
Registered: March 2013

Senior Member

Hey,

I haven't used the MICS (would like to someday), so I can't offer much. There is a discussion of the weights that you can download, but it focuses on how they are computed, not how they are used. That said, this bit seems relevant:

The procedure for applying the post-stratification weights depends on whether (i) weighting is done at the level of each sample person (or household) before the computation of any aggregate measures such as separate indicators, or (ii) if separate indicators (such as by urban-rural or region) are computed first and then put together with appropriate weights....
All indicators, whether at the national or domain level, and whether applicable to the whole population or to specific subgroups (such as children of a particular age or gender), automatically incorporate the post-stratification adjustment since these weights have been applied at the level of the individual cases

So that sounds like you can use the same weights - but I might be totally misreading that. Here is the link to the documentation - download the "Sampling Details" section. I've attached it anyway. Hope it helps - let me know if you find anything interesting.