The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Countries » Bangladesh » Zero values for weight variable?
Zero values for weight variable? Mon, 16 February 2015 20:28
 mmr-UMICH Messages: 21Registered: February 2015 Location: A2, MI Member
Dear DHS Forum:

I am working on women's flat data for 2011 of Bangladesh for some sort of multilevel analysis and wanted to prepare weight variables for both the individual women (clusters in HH level=v002) and HHs (clusters in PSU level=v001) levels. I found that 93 cases have 0 for weight variable, v005. I prepared a file for such cases plus matching key values for V001 and v002 and for your information (FYI), first few lines are given below:

[sorted and merged two files]
v001 v002 v005 #of women
------------------------------------
6 69 0 1 <------(1)
25 126 0 1 <----- (2)
25 126 275954 1 <------(3)
29 52 0 1
29 52 664763 2
30 93 0 1
32 50 0 1
32 50 319113 5
[many lines omitted ...]

Since statistical software will discard those cases for any kind weighted analysis, so we have to find weight values for those women. Based on the above data, it may be clear we can assign 275954 to v005 for eqn(2) from egn(3) because they are from the same cluster and HH. I am giving some detail info below. For eqn(1) (cluster and HH combination is only single woman interviewed) and eqn(4) below, I saw that all weights are the same for cluster 6 (=v001) and seemed to me we can also assign 609991 to v005 for women in cluter 6 and HH = 69. Do you think the process of replacing 0 weights correct? FYI, there is no zero weights for HHs.

[sorted by v001 v002 v005 for uniques]
v001 v002 v005 #of women
--------------------------------
[lines omitted ...]
6 2 609991 1
6 7 609991 1
6 11 609991 1
6 19 609991 1
6 27 609991 1
6 31 609991 1
6 35 609991 1
6 44 609991 1
6 48 609991 1
6 56 609991 1
6 60 609991 1
6 64 609991 1
6 69 0 1 <---- (4)
6 73 609991 1
6 77 609991 2
6 81 609991 3
6 85 609991 1
6 89 609991 2
6 93 609991 1
6 97 609991 1
6 102 609991 1
6 106 609991 1
6 110 609991 1
6 114 609991 1
6 122 609991 1
[lines omitted ...]

Re: Zero values for weight variable? [message #3826 is a reply to message #3808] Wed, 18 February 2015 13:19
 mmr-UMICH Messages: 21Registered: February 2015 Location: A2, MI Member
Hello DHS team:

I had a question (please see message # 3808) regarding sampling weight variable, v005, of women's flat file of Bangladesh DHS 2011. There are 93 cases with 0 for v005 variable. Could any one from DHS team help me in this matter at your conveniences?

Thanking you.

Moshiur
Re: Zero values for weight variable? [message #3838 is a reply to message #3826] Sat, 21 February 2015 16:07
 Trevor-DHS Messages: 680Registered: January 2013 Senior Member
These cases are all women under the age of 15 and are dropped from the analysis. The analysis weights are all calculated based on women 15-49 only. You are correct that all women in the same cluster have the same weight and if you wish to you can use this weight for the 13 and 14 year olds, but note that the adjustments to the weights for non-response if 13 and 14 year olds were included would be slightly different than for 15-49 year olds. We do not recommend including the 13 and 14 year olds in analysis.
 Previous Topic: variables for svyset in Stata (Bangladesh 2011) Next Topic: Wealth Quintile calculation
Goto Forum:

Current Time: Sun Mar 29 14:46:06 Eastern Daylight Time 2020