Interpretation of weights for scaling to national level [message #10572] |
Wed, 10 August 2016 18:19 |
SPA_question
Messages: 2 Registered: August 2016
|
Member |
|
|
If I am trying to scale counts reported by facilities in the SPA survey up to national levels (e.g. number of patients diagnosed with KAPOSI'S SARCOMA in the last month (KEN 2010 SPA)), how do I interpret the facility weights? I see people are dividing the weights by 1,000,000... is it really that simple? Thanks!
|
|
|
Re: Interpretation of weights for scaling to national level [message #10598 is a reply to message #10572] |
Mon, 15 August 2016 10:19 |
Liz-DHS
Messages: 1516 Registered: February 2013
|
Senior Member |
|
|
Dear User,
A response from Dr. Ruilin Ren, sampling expert:
Quote:
All the DHS surveys use normalized weights which is re-scaled to be a unit weight, that is, the number of weighted cases and unweighted cases are identical at national level. A normalized weight is a relative weight (the scale of the weight was removed), so you cannot use it to produce estimation of totals. The weighted total just give you the number of cases interviewed which is the effective sample size. It is true that when you do analysis with DHS data, you need to divide the weight variable by 1,000,000 because the weight variable was multiplied by 1,000,000 in the recode during data processing to achieve integer numbers. But this process has nothing to do with re-scaling.
|
|
|
|
Re: Interpretation of weights for scaling to national level [message #10738 is a reply to message #10618] |
Wed, 07 September 2016 08:58 |
Liz-DHS
Messages: 1516 Registered: February 2013
|
Senior Member |
|
|
A response from senior sampler, Dr. Ruilin Ren:
Quote:
The answer for the question is yes for overall national weighted averages multiply by the total number of facilities in the country to get an estimated national total. But it will not be valid for any restricted averages because the weight is a relative weight.
Thanks
|
|
|