Normal user
8 May 2015, 11:46
1. how do you perform a weighted analysis for anthro results? ( manually)
2.what are parameters to consider in case you want to conduct 5 separate SMART surveys within a County.
3.suppose we have a county estimate(s) for the 5 separate surveys, does design effect have any effect on result anthro (s) interpretation?what are key factors to consider when interpreting county estimates from weighted analysis of the 5 separate assessments?
Frequent user
10 May 2015, 08:26
(1) I think this is covered in
this post. Post again here if this is not what you need.
(2) I am not sure what you mean. To combine surveys you need population in each survey area to use as weights.
(3)(a) Design effects need to be accounted for. The method outlined in
this post does this by calculation standard errors from the 95% CIs on estimates. You need to be sure that the 95% CIs were calculated for the survey design (ENA for SMART will do this). Pooling SEs from each survey estimate will account for the design effect at the higher (i.e. combined surveys) level.
(3)(b) You need to be sure that the overall estimate is a sensible thing to quote. This means that the estimates you are combining as similar. If not you will hide variability.
I hope this is of some use.
Normal user
10 May 2015, 18:10
Thanks Mark
very much helpful!