ReSampling for representativeness

[SASLeaner336]

Greetings all,

I have a sample which appeared to be not representative and I am trying to resample it so it becomes representative of my target population. I have searched for many days but I am still stuck. If anyone has an idea I would be grateful.

Thanks.

[ballardw]

First you might describe some details such as 1) size of the original data , 2) the size of the sample, 3) how you drew the sample and 4) what test(s) you ran and the results that make you think the sample is not "representative".

Perhaps your "sample" was generated incorrectly such as a simple random sample of xx records when your data actually has strata that you have reason to believe have different characteristics such as high/low account worth; income; geography; or some other combinations of characteristics.

And sometimes you just don't have data to match a "target population". If my data is from California and my target population is in New Zealand I might not expect a good match.

[SASLeaner336]

My data is about prepaid Telecom users so you will find variables like the quantity of data used, ARPU, the lifetime of the user...

My sample's size is : 237707 users

My target population's size is : 1200803 users

I drew the sample using proc surveyselect with the method SRS.

I ran the khi2 test on my qualitative variables and TTEST on my quantitative variables and my results were significant. But when I compared the mean of the ARPU of my sample to the mean of the ARPU of my target population I found that they were significatly different. That's why my sample isn't representative.

I can't redo a sampling on my whole population because some action was performed on my target population. That's why the only change I can do is on the sample.

I don't think that I have the last problem since they all come from the same population.

Thank you for your response.