Greetings,
It would be great to know your thoughts on the following question which I received from a friend:
"the results of the recent household survey show that 3.5% of the anthropometry data were extreme WAZ values (outside the range of -6 to +5 SD).
We are checking to see if these extreme values are related to 1 or 2 enumerators, or if these values are equally dispersed among all enumerators.
I found one reference that spoke to 3% being high (p49 of http://siteresources.worldbank.org/INTPAH/Resources/Publications/459843-1195594469249/HealthEquityCh4.pdf)
I can’t seem to find any literature/ standard recommendations for quality of data regarding extreme values.
Yes, we will exclude the extreme values (the program filters these out automatically), but is the overall quality of the survey in question if there are these many extreme values?"
Many thanks
Hello,
To see if you can trust your data collection, you can look at WHZ values: do you have also a lot of extreme or aberrant values?
WAZ is influenced by age which is one of the most difficult data to collect properly: maybe you should investigate how the age was collected (events calendar, cross questions...), does this population have problems to remember age of children (especially the older ones)?
Your survey's data might be reliable for acute malnutrition results but less for chronic malnutrition and underweight results if the problem comes from age data collection.
Hope this help
Pascale Delchevalerie
Answered:
12 years ago