The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Dataset use in Stata » Family Planning in young women (STATA use)
Family Planning in young women [message #17974] Mon, 05 August 2019 20:08 Go to next message
Kyaw is currently offline  Kyaw
Messages: 2
Registered: August 2019
I am using the Myanmar DHS 2015-16 data to identify the determinants of contraceptive use among 15-19 and 20-24 married women. The observations selected was "15-19 = 235" and "20-24 = 859".
Svyset was used. The independent variables are the socio-demographic and other variables like exposure to family planning messages. The dependent variable is "modern contraceptive use".

On using the "svy, subpop ( ), tab" command to analyze the two age groups separately, some of the cells become either 0 or lower than 5.

Variable --> v013 - 15-19 = 1
- 20-24 = 2

1. The cells with 0 observations showed a message "Table contains a zero in the marginals. Statistics cannot be computed". I used the command "svy if v013==1, tab" command and the 0 disappeared. The proportions and confidence interval are the same but I got the p-value. However, is it recommended to use this command or is there an alternative?

2. Some cells have observations less than 5. Usually Fisher's exact test is used in very small numbers but in this survey set, it is not compatible (I think). Is there a way to calculate them?

best regards
Re: Family Planning in young women [message #18014 is a reply to message #17974] Fri, 16 August 2019 14:23 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 1965
Registered: February 2013
Senior Member
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:

It is definitely preferable to use the subpop option. However, in your situation I would use these lines:

gen v013_1=1 if v013==1
gen v013_2=1 if v013==2
svy, subpop(v013_1), tab
svy, subpop(v013_2), tab

I think this will take care of the problem. In the specification of subpop you need to use an argument (what goes between the parentheses) that is 1 for v013=1 or for v013=2.
Previous Topic: Duplicated label categories
Next Topic: 2017 Ghana Maternal Health Survey
Goto Forum:

Current Time: Sat Oct 24 11:03:28 Eastern Daylight Time 2020