Selection of Regions In Pakistan (PDHS 2017-18) & Sample Weight [message #20330] |
Fri, 23 October 2020 03:24 |
shujaat.smc@gmail.com
Messages: 75 Registered: July 2020
|
Senior Member |
|
|
Dear DHS Representative,
I am working on BR-file (PDHS 2017-18) for determining the predictors of neonatal mortality in Pakistan.
In PDHS 2017-18 there are Eight Regions (Punjab, Sindh, KPK, Balochistan, ICT, Azad Jammu & Kashmir, Gilgit Baltistan & FATA.
I am using v005/1000000 as my sample weight as per your DHS tutorial on Youtube.
Q1: I want your confirmation that though it's clear that Azad Jammu & Kashmir & Gilgit Baltistan have different weights (v005s) therefore for my analysis, I have to take 6 regions of Pakistan excluding AJK & GB?
Q2 What will happen to my analysis (using survival analysis) If I select only four regions (Punjab, Sindh, KPK & Balochistan)?
{After watching the DHS video on Youtube "Weight or Not to Weight" I think its legal obligation on me on basis of the statistical principle that I must do my analysis on six regions of Pakistan that are Punjab, Sindh, KPK, Balochistan, ICT & FATA)
Waiting for your kind & elaborate reply.
Best Regards
Dr. Hussain
|
|
|
Re: Selection of Regions In Pakistan (PDHS 2017-18) & Sample Weight [message #20364 is a reply to message #20330] |
Mon, 26 October 2020 09:50 |
Bridgette-DHS
Messages: 3195 Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Lead Statistician, Ruilin Ren:
Q1. Yes, the two special regions AJK and GB have different weights. These two regions are standalone, they cannot be combined with any other region because of their normalized weights. You can do analysis for the Pakistan (6 regions, without AJK and GB), AJK, and GB separately
Q2. Yes, you can restrict your analysis to the 4 major provinces Punjab, Sindh, KPK & Balochistan. But you need to let the reader know that your analysis include only the 4 major provinces. I do not think that in the "Weight or not Weight" video we stated that you must do analysis for a country as a whole. You can create your study domains by grouping regions, but not by dividing a region, because the sample size will be too small in the latter case.
|
|
|
|