Home » Countries » Bangladesh » impact of family composition on unmet need for family planning in Pakistan (child parity with child sex respect to birth order)
impact of family composition on unmet need for family planning in Pakistan [message #21630] |
Tue, 01 December 2020 04:30 |
Rubab Saleem
Messages: 6 Registered: November 2020
|
Member |
|
|
hello everyone,
my thesis title is "impact of family composition on unmet need for family planning in Pakistan". I am working on PDHS 2017-18. I am trying to generate the child parity variable with respect to child birth order using (BORD) for birth order and b4_01 for sex of child and b5_01 for child is alive or single birth. once I have generated this variable till the birth order5, in the regression the values from the 4th birth order are missing no values shows in regression and all the values exist in 5th parity group. kindly suggest me how I can generate the accurate the variable and on other condition how I an use the just birth order (BORD) for my thesis analysis. I have attached my do file for parity for better understanding and reference paper. thanks A lot. waiting for your response.
Rubab Saleem
Pakistan
|
|
|
Re: impact of family composition on unmet need for family planning in Pakistan [message #21638 is a reply to message #21630] |
Wed, 02 December 2020 08:38 |
Bridgette-DHS
Messages: 3169 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
I don't understand what you mean by "trying to generate the child parity variable with respect to child birth order". The child's birth order is given by bord, as you say. Parity is a characteristic of the mother, not the child. It is v201 at the time of the survey; before then, it is the same as the birth order of her most recent birth. The reverse numbering of births, beginning with 1 for the most recent, is bidx, which is the subscript in the IR file. If bord is NA (not applicable, i.e. a dot) then the woman had no child of that birth order. The birth history variables in the IR file are constructed for up to 20 births but the great majority of them are empty (NA).
Since you are looking at household composition, perhaps you want to restrict to children who are alive and living with the mother. You can identify those children with the PR file (hv112 may help). Also b9 identifies whether the child is alive and living with the mother.
If you still have a question, please clarify what you want to do.
|
|
|
|
|
|
|
Re: impact of family composition on unmet need for family planning in Pakistan [message #21830 is a reply to message #21827] |
Mon, 28 December 2020 08:04 |
Bridgette-DHS
Messages: 3169 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
We recently completed a report (Analytical Study 76, https://www.dhsprogram.com/pubs/pdf/AS76/AS76.pdf) that is related to what you want to do. I also suggest the following. You could calculate for each cluster the proportion of households (or individuals) who are in the bottom two wealth quintiles (or in the top two, or above the median value of the continuous wealth index, v191 or hv271, etc.).
For example, you could construct a binary individual-level variable and then use egen (mean) to construct a cluster-level mean or proportion, as follows:
gen WI_low_ind=0
replace WI_low_ind=1 if v190<=2
egen WI_low_cl=mean(WI_low_ind), by(v001)
Here, the cluster-level ("cl") variable would be the same for every woman in the same cluster--it's the proportion of women in the cluster whose households are in the bottom two wealth quintiles. You could reverse the 0 and 1 so that "1" would refer to higher quintiles rather than lower quintiles. You could also construct it with households, rather than individual women, as the units of analysis, but that would require using the HR or PR files and then merging with the IR file.
If you calculate multiple cluster-level indicators--for example, the proportion of women in each quintile in each cluster, and put them all into a regression, it will be hard to interpret the results. Good luck!
|
|
|
|
Re: impact of family composition on unmet need for family planning in Pakistan [message #22067 is a reply to message #22066] |
Tue, 26 January 2021 15:00 |
Bridgette-DHS
Messages: 3169 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
The file to use is determined by what is the dependent variable. In the IR file, women are the units. The number of births in the past five years is v208. In the BR file, the children of those women are the units, and b11 is the length of the preceding birth interval. If the birth is a first birth (bord=1), then b11 is a dot (.), for not applicable. That is, the interval is only defined for births after the first birth. You could define a binary variable to be 1 if b11<24 and 0 otherwise, among non-first births (bord>1). Then use logit regression. Predictors would come from the mother's data, which is already attached to the child's records.
|
|
|
|
Re: impact of family composition on unmet need for family planning in Pakistan [message #23016 is a reply to message #23014] |
Fri, 25 June 2021 07:31 |
Bridgette-DHS
Messages: 3169 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
For the first question--there is no way to produce any indicators for all Pakistan including provinces 5 and 7. This was the decision of the government of Pakistan. We cannot help you to do something that the country explicitly rejects.
For the second question--there have been many postings about weights on the forum. Yes, you should use weights (pweights for estimation commands). Weights are case-specific, not variable-specific. You should prepare an svyset command that also adjusts for weighting and stratification, and use "svy: " in front of estimation commands.
Please review what has already been posted.
|
|
|
Goto Forum:
Current Time: Mon Oct 7 16:35:54 Coordinated Universal Time 2024
|