(I'm not planning to do any analysis of change over time, so the time period covered doesn't matter to me.)

I have a few questions.

1. Can you point me to some more complete/technical documentation about the sampling strategy and what samples are representative at what level?

2. How should the weights be treated when pooling multiple years? Does the pooled download (2015 & 2016) include appropriate weights for the pooled sample, or do they need to be adjusted somehow?

3. I see from reading the forums that the wealth quintiles need to be adjusted when pooling data (since they are quintiled relative to the individual year samples). Does the pooled download include wealth quintiles for the pooled sample or within-sample quintiles?

Thanks

Sarah]]>

Following is a response from Senior Sampling Statistician, Ruilin Ren:

The Senegal continuous DHS had been modified a little in the course of the survey to incorporate needed changes. The document you cited reflects the initial design of the survey but did not integrate the modifications in the course of the survey. Actually the survey was continuous from 2013 to 2016, with the 2013 and 2014 surveys based on previous population census (2002), while the 2015 and 2016 survey were based on the new population census (2013). The main difference was the definition of urban-rural. So it is advised that you can combine the 2013-2014 surveys together, and 2015-2016 surveys together. ICF already released the pooled data sets with combined weights calculated for 2013-2014 and 2015-2016, so you do not need to do anything on the weights if you use the pooled data released by ICF. However, the 2017 survey was a standalone survey with designed representative results for each of the 14 regions.

As for the wealth quintile for the pooled data, I only know it was not recalculated based on the pooled data, but through a specific procedure. I do not have all the details. If you need the details, please let us know and we will direct the question to the appropriate expert from ICF.

]]>

I am planning to use Peru continuous DHS. Which survey years I need to combine since 2004? After I put together the datasets, is it necessary to recalculte the weights?

Thank you so much,

Ana Elisa ]]>

Each survey year can be used individually since each is nationally representative. For reduced sampling error 3-year moving combined surveys can be created through 2008 using un-normalized weights. Surveys after 2008 are large enough for reduced sampling errors.

]]>