Hi,
Anyone working with calculating sampling weights for complex sampling designs?
If you have a sampling stage where you use stratified sampling to recruit the participants from a cluster and you decide the size of each strata based on the national population, do you still need to use the sampling weights for that stage?
Much appreciated your views.
Thanks.
Deepani
if you want to make a comparative between your samples based on different type. and you should take the sample of every sector/type based on the percentage of the sector in the original population after that you take the sample based on that percentage.
Please, if you still need example or if you want me to help you to calculate your sample, you can contact with me.
Deepani -
The probabilities of selection at each stage matter.
I do not know what you mean by saying you "...decide the size of each strata based on the national population...." If you mean some kind of proportional allocation, that is not very efficient. Allocation between strata should depend upon sigma in each stratum. But even if you used proportional allocation within each cluster, the national population is removed from that. Once you go to a second stage, survey weights become more complicated. It is complicated enough with simple random sampling at each stage. For one stage, stratification is often very helpful. Perhaps you are confusing the terms "stratum," and "cluster." ??
You might want to research the terms "two-stage sampling," and "three-stage sampling." The different stages have probabilities of selection based first on the N clusters, and then the Mi members of cluster i, and perhaps down a further level/stage. Stratification for each cluster, rather than simple random sampling, would mean a more complicated probability of selection within a stage also.
Perhaps you want to first make certain that you understand two-stage sampling with simple random sampling each:
https://onlinecourses.science.psu.edu/stat506/node/44
A good textbook could also be very helpful. Here are two suggestions:
Thompson, S.K.(2012), Sampling, 3rd ed, John Wiley & Sons.
Cochran, W.G(1977), Sampling Techniques, 3rd ed., John Wiley & Sons.
Perhaps I have misunderstood your question, and you know all of the above, but this was my guess as to what you need to consider.
If you just want to know about stratified random sampling, those are good books for that as well, and there is this:
https://onlinecourses.science.psu.edu/stat506/node/27
Best wishes - Jim
Hi Deepani, I agree with Prof Knaub. His comments and references contain all you need to know to carry out your research. You do not need to keep going back to your original sample space or population at a lower stage of your multistage sampling. That would amount to multiple counting. I hope my comment further amplifies Prof Knaub's advice.
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Sampling Design