| Average Mosquito net age [message #2694] |
Mon, 11 August 2014 06:19  |
|
|
Dear DHS,
A quick one, i'm looking at the mean/median mosquito net age. I'm thinking it a reshaped long variable HML4 to use. However, i realise all nets older than 36 months were grouped into 1 category 96.
How do i get the mean/median age in this case?
Also, for categorical variables which include the options "Don't know",Do you always include them in the denominator or not? i'm thinking including them could under estimate the indicator if the don't knows are significantly large.
Thanks
Geoffrey
|
|
|
|
| Re: Average Mosquito net age [message #2720 is a reply to message #2694] |
Fri, 15 August 2014 01:27  |
Trevor-DHS
Messages: 789
Registered: January 2013
|
Senior Member |
|
|
You won't be able to calculate a mean as anything longer than 36 months is coded 96, but you can still calculate a median. Recode 96 to 36, and then calculate the median excluding all of the other special values (98, 99, etc.). In fact if there are only a few cases coded 96, then you can come up with something close to the mean in the same way as the median. It won't be exactly correct because those longer than 36 months are effectively truncated at 36 months, but it won't be far off if there are only a few cases coded 96 originally.
For many categorical variables, we do include the 98s in the denominator as these are often similar to "No" responses and we are usually interested in the "Yes" responses, and the small number that we would exclude would have little effect on the result. Including them in the denominator may slightly under estimate the indicator, but excluding them from the denominator may slightly overestimate the denominator. However, if there are many Don't know responses then we would typically exclude them from both numerator and denominator as they would clearly bias the estimate, but we would typically note that this estimate was based on excluding the Don't knows.
|
|
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
