Menu ENN Search
Language: English Français

Weight factor?

This question was posted the Assessment forum area and has 2 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Assaye

Normal user

21 Dec 2010, 10:47

Hi all,

I have done a nutrition survey in 3 districts under a region and I want to know the average for the region.

I think I have to calculate weight factor for each districts to get the regional level average. Would you please show me how to calculate weight factor?

Thanks a lot!

Rossella

World food programme

Normal user

21 Dec 2010, 13:50

Hi,

if you get from this link the WFP Comprehensive Food security and Vulnerability analisys guidelines, you can find in chapter 4 a detailed description on how to calculate weight factor.
Hope it helps.

http://www.wfp.org/food-security/guidelines

Mark Myatt

Consultant Epideomiologist

Frequent user

21 Dec 2010, 14:49

The usual approach is to take a weighted average.

To do this you need to have some idea of the (relative) population of each district. Example:

Prevalences:

District 1 : 7.6%
District 2 : 9.3%
District 3 : 5.4%

If you have populations :

District 1 : 123,000
District 2 : 91,000
District 3 : 52,000

Then the regional average is:

(7.6 * 123000 + 9.3 * 91000 + 5.4 * 52000) / (123000 + 91000 + 52000) = 7.75%

It is best to have population data but coarse weights based on relative sizes can work. Using the example above we might use small (1), medium (2), large (3) as coarse weights:

p = (7.6 * 3 + 9.3 * 2 + 5.4 * 1) / (3 + 2 + 1) = 7.8%

There are alternative approaches to weighting but they are usually, at bottom, just alternative ways of taking a weighted average. You may need to use these alternative methods with stats packages.

The tricky bit is in calculating a confidence interval. This may be more difficult in your case as you already have a complex (i.e. cluster) sample at the district level. I suggest that you do this with software that can work with such a complex sample. A "rough and ready" approach would be to use a weighted average of the upper and lower 95% confidence limits from the district level samples. This is very much an approximate method.

I hope this helps.

Back to top

» Post a reply