Home » Topics » Reproductive Health » Contraception discontinuation
| Re: Contraception discontinuation [message #9602 is a reply to message #9552] |
Wed, 20 April 2016 14:19   |
Trevor-DHS
Messages: 802
Registered: January 2013
|
Senior Member |
|
|
I don't have too much experience with life tables in Stata, but analysis expert Sarah Bradley wrote recently for another user:
"My suggestions for replicating DHS results are:
1. Reshape vcal_1 and vcal_2 into files where the unit of analysis is the episode of contraceptive use (potential multiple episodes j per woman i). It sounds like this isn't a daunting task for you, but if you run into trouble with this step, please let Trevor and I know, as we're working on standard events-based files.
2. Calculate durations of use from the start and end of each episode of use. Note that DHS standard is to use a window of observation that is 3-62 months prior to interview, censoring the most recent 3 months prior to interview, dropping left-truncated episodes, and including episodes of use that began >62 months before interview as late entries into the life table. We describe this in somewhat excruciating detail in the methodological appendix in the recent Guttmacher Report on Contraceptive Failure.
3. For associated-single decrement rates (as in the Guttmacher report), try -sts list-. Note that this doesn't support pweights. I use iweights and calculate the SEs in alternative ways.
4. For multiple-decrement rates (as in DHS final reports, STATcompiler), I like -stcompet-. The documentation on this in Stata Journal is quite good.
A caveat: -ltable- in Stata assumes that nax = n/2, i.e. that discontinuations occur in the middle of the month. The DHS standard is to assume that discontinuations occur at the end of the month, so -ltable- will produce slightly different estimates than -sts list- which assumes that failures occur at the end of the month."
|
|
|
|
Current Time: Thu Oct 10 06:18:24 Coordinated Universal Time 2024
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
