Question on weighting of survey data

#1
Hi there,

I have recently concluded a survey and am working on the weighting of the survey data. I would like to weight the survey data by weekend vs weekday combination weights. However, looking at the data, I realised that some groups that I would like to analyse the data by, were not surveyed on weekends. Hence, I do not have weekend samples for some of the groups. My questions are:
1) Can I weight the data by weekend vs weekday if some groups do not have a weekend sample?
2) If yes to 1), how should I treat the groups that do not have a weekend sample?
3) If no to 1), how should I weight the data?

Population figures are available by Total and weekend vs weekday splits for all groups.

Would appreciate very much any help that can be provided.

Thanks!
 

noetsi

No cake for spunky
#2
Why are you weighting the data? And why don't you have a weekend/weekday sample for some.

If every group is important to you and you don't have the data to generate weights for some I don't see how you could possibly generates weights. Whatever method/software (to make only one point) is going to want to know what weight you have assigned them.

I think you either have to exclude some groups from the analysis or not weight the data for that analysis.
 
#3
Thanks, Noetsi. Need to weight the data as we need to analyse the results by weighted data as well. We have weekday sample for all groups but realised post survey that some groups were not surveyed on weekends (when they should have been surveyed) hence there is no weekend sample for these groups. Yes, all groups are important so we need to decide the correct method for weighting the data. Under the circumstances, would it be correct to:

a) generate weekday/weekend weights and combine the groups (that do not have weekend sample) with another group/groups? OR

b) generate weekday/weekend weights but drop the groups that do not have a weekend sample OR

c) generate weights based only on total population for each group?

From your response it appears that either b) or c) would be appropriate......is a) a viable/statistically valid option at all......?

Thanks again for your interest and response.
 
#4
Thanks for discussing such an important statistical terminologies.This thread at a good source of knowledge enhancement about the statistical terminologies.








HCG Drops
 
Last edited:
#5
Generally, I will reply on questions realting to statistical test chice or study design and others related to biological applied stats. I tend to leave PDF questions and equation based questions alone.



--------------------------
The most attractive feature of designer Lingerie Dress very cheap is its soft and delicate embroidery designs.