Weighting for calendar module [message #29411] |
Wed, 12 June 2024 19:03 ![Go to next message Go to next message](/theme/default/images/down.png) |
seinkim
Messages: 4 Registered: October 2022
|
Member |
|
|
Dear DHS team,
I'm trying to analyize trends in using different contraceptive methods in Zambia using a calendar module as well as sub-group analysis, such as mean rate of contraceptive methods comparing urban to rural (or teeangers vs adults).
The data structure is like below:
ID month-year condom
1. Jan-2013. 1
1. Feb-2013. 0
1. Mar-2013. 0
.
.
1 Dec-2013. 1
ID
2. Jan-2013. 1
2. Feb-2013. 1
2. Mar-2013. 1
.
.
2 Dec-2013. 1
I want to calculate aggregated rate of using contraceptives each month, and each months have different numbers of observations. (not all individuals followed up for same lenght of months)
To do this, I used following code:
collapse (mean) condom injectable f_steralization m_steralization implants iud traditional [pweight = weight] , by(cmc teen )
Using this numbers, graphs were produced and I did regression analysis. I think I have to use individual weights but I'm not sure whether I'm doing correctly. Also, if I want to do sub-group analysis by urban vs rural, should I apply individual weights to calculate the aggregated mean by sub-groups?
Thank you!
|
|
|
Re: Weighting for calendar module [message #29417 is a reply to message #29411] |
Thu, 13 June 2024 15:03 ![Go to previous message Go to previous message](/theme/default/images/up.png) |
Bridgette-DHS
Messages: 3151 Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS staff member, Tom Pullum:
Your use of weights within the collapse command is sufficient for the aggregation. For analysis with urban/rural groups you would include v025 in the by() option.
Normally with collapse I use "[iweight=v005/1000000]" but it should be equivalent to what you did. In the aggregation there is no adjustment for the cluster or stratum effects that are part of the full svyset command but for a descriptive analysis that's ok. Weights are the only component of svyset that affect point estimates such as means and rates.
|
|
|