What is this technique, and what software might support it?

#1
Hi

I'm not a statistician, but would really appreciate some help in resolving a problem I have with producing a representative sample for a mail survey (I hope that I've put this in the right forum - it was a best guess). I'll explain my problem with an example:

Consider a deck of standard playing cards. From this deck I want to select a random sample that meets the following criteria:

a) 10 cards total
b) 5 red, 5 black
c) 2 cards must be a Jack, Queen, or King (e.g. could be 2 Queens, a Queen and a King, etc.)
d) Sample cannot include 6 of spades, 7 of clubs, 7 of hearts, or 10 of clubs.

This problem started out as simple stratified sampling (condition b only), but became more complicated with the addition of more requirements to make the sample representative of multiple strata simultaneously. My actual problem is much the same concept, but more complicated (population of approx 5000, sample approx 1000, approx 10% must be excluded from the sample for specific reasons, approx 30 conditions).

My questions are:

1) Is this still stratified sample? If not, what is it (if it has a name)?
2) Could anyone enlighten me as to which software products are able to undertake this sampling?

Any help appreciated.

Thanks

Stuart
 

noetsi

Fortran must die
#2
You are asking a sampling question more than a statistical question. For the most part statistical software stresses statistics rather than sampling. You can use SAS (and I assume the other software) to make a random or stratified random sample, but it won't suggest to you what exactly you need to sample or how to do it. I suspect in practice you would have to set up the table to do this before you sampled (for example excluding cards you don't want and generating tables that have only face cards from which you randomly sample to ensure you have enough of each of these types). This could be done either manually or through SQL.

If you do this and you use regression, for example, to analyze any results you will have to counterweight to address the stratification.
 
#3
Thanks for that response - if any modorator thinks that this should be moved to a more appropriate thread, please do so.

Any other thoughts from the community?

Stuart