Dear Allstat
Forgive my potentially non-expertise in this field. I have just received an "analysed" dataset back from a UK agency which shows responses from a national questionnaire expressed as simple freqencies and percentage response rates for each question. Running the same questionnaire data in a standard stats package gives results that are on average 10 - 18% lower for each question. The reason is that the "official analysis" has discounted all the missing data (non-responders) and simply expressed the results as a percentage of the total numbers who responded to each question.
Is this correct or are there issues around how the missing data should be accounted/corrected for in simple expressions of data like this?
On the surface, seems like a dumb question to ask but I can't find a straight answer only that there are various approaches to deal with it, concerns etc etc
Brian
An NHS Trust
|