Title of the project: To determine the awareness of antimicrobial resistance (AMR) in population from F-north ward Mumbai, India.
What is known:
Studies have shown that AMR awareness in India is 3%
F North ward Mumbai population = 700,000
Sample size calculation.
Estimated Proportion 0.3
Desired precision of estimate 0.05
Confidence level 0.95
Population size N/A
Sample size 323
Should I go for cluster sampling? How do I decide how small clusters should be? or How many people in one cluster?
Please clarify your question. You say AMR awareness is 3%, but then you say the estimated proportion is 0.3, which would be 30%. Which is correct? If you meant the proportion is 0.03, did you also mean the desired precision is .005 (not 0.05)?
In general, avoid cluster sampling if you can. If you have a sampling frame that allows you to select cases directly (e.g., individual adults) rather than in clusters (e.g., dwelling units), that will give you better precision.
If you must cluster, sample as many clusters as your research budget can afford, and the fewest affordable number of people per cluster. In the extreme, if you could randomly sample 323 clusters and then sample only one person from each cluster, then the clustering shouldn't adversely affect the precision of your estimates much if any.
If you have no choice but to cluster, large clusters are generally better than small clusters, because large clusters are likely to be more diverse (greater variances on the variables of interest) and therefore more similar to the target population as a whole. With small, homogeneous clusters, each additional person you sample from a cluster essentially just repeats the information about the population you get from the first person in that cluster. So in that case, the precision of your estimates will be much worse than the nominal sample size (total # of people) would suggest.
proportion is 0.03 not 0.3 and dont understand where is the need for cluster sampling while you have the frame (I think census report is there). Try use simple random sampling instead and to determine the sample size use the sample size calculator at the http://sample.iasri.res.in/ssrs and there is a sample size calculator. Use it and go for simple random sampling without replacement.
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Sampling Design