I have a most at risk population (MARP) of about 8,013 with sampling frames as follows.
Strata 1. MSM – 1674
Strata 2: FSW – 1977
Strata 3: BB- 737
Strata 4: DU-3625
Total = 8,013
If I calculate total sample size with 95% CI, 50% assumed proportion, and acceptable difference as 5% and total population size as 8,013. The sample size would come as 367. With 10% refusal It will be 404.
My question is after stratified random sampling in this scenario, Is it possible me to analyse the strata data independently (MSM, FSW, BB, or DU separately) and generalize the values to the strata?
What would you recommend If I want to analyse the groups independently.
Karawita -
First, are your data all yes/no data because I have not checked but it looks like you used one of those online 'calculators' that assumes yes/no data? They also assume simple random sampling, and usually are for the worst case where p=q=0.5, and ignore the finite population correction factor.
As for stratification, it looks like you are really trying to publish for different categories, not use stratification to improve an overall answer, so you have to think of each as its own independent simple random sample, if i read your question correctly.
The key to sample size is the standard deviation for each population. Cochran (below) has some suggestions for guessing this, including a pilot study.
For your purposes, I suggest you check a textbook or two, such as one of these:
Cochran, W.G(1977), Sampling Techniques, 3rd ed., John Wiley & Sons.
Blair, E. and Blair, J(2015), Applied Survey Sampling, Sage Publications.
Lohr, S.L(2010), Sampling: Design and Analysis, 2nd ed., Brooks/Cole.
Cheers - Jim
Karawita -
First, are your data all yes/no data because I have not checked but it looks like you used one of those online 'calculators' that assumes yes/no data? They also assume simple random sampling, and usually are for the worst case where p=q=0.5, and ignore the finite population correction factor.
As for stratification, it looks like you are really trying to publish for different categories, not use stratification to improve an overall answer, so you have to think of each as its own independent simple random sample, if i read your question correctly.
The key to sample size is the standard deviation for each population. Cochran (below) has some suggestions for guessing this, including a pilot study.
For your purposes, I suggest you check a textbook or two, such as one of these:
Cochran, W.G(1977), Sampling Techniques, 3rd ed., John Wiley & Sons.
Blair, E. and Blair, J(2015), Applied Survey Sampling, Sage Publications.
Lohr, S.L(2010), Sampling: Design and Analysis, 2nd ed., Brooks/Cole.
Cheers - Jim
Dear Dr. Karawita,
I also think you need to treat each category separately (do SRM for each category separately), if you need to analyze proportions for MSM, FSW, BB, DU separately and compare the values. So you may need a higher sample size (around four times) than you mentioned.
You can come up with the proportion of successes / Failures of whole group (not for each category), if you pool all four categories and do SRM (As you have mentioned).
Thank you,
Suchira
Dear Dr. Karavita,
You can do analysis in seperate strata. However combining the strata data to give an overall view will require the use of sampling weights.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size