Dear all
I hope to fine the full answering for my two questions
1- If the population of a city is 800,000 and the sample size (n) is 450 what is the pilot study sample size and if I took pilot study sample size is 60 is true or not on my consideration I think the sample size of any pilot study should not less than 10% of the total sample size so if the sample size is 450 --------------------------- pilot sample size is 45 ----------- X*100/450=10%
10*450=100*X------------------------------------ n (pilot study)= 45
2- If I have Sample Size for different income levels (high, middle and low)
and If high level is 150 & middle level is 160 & low level 90
and the different districts in the different levels are high levels 20 & middle level 35 & low level 25
so please how i can find a relation between the different income districts and participants in the different income levels
Hi! According to central limit theorem using minimum sample size of 30, data tends to get normal for pilot study. Thanks
Hello Ossama,
The size for a pilot study should depend on what it is you are trying to find out from the pilot study. If it is to check on survey procedures, logistics, how to elicit responses, and so forth, then a relatively small sample can suffice (perhaps 30 cases). If you are trying to estimate population values in order to get a more defensible estimate of the full study sample size, then 60 might be adequate (more would be better, of course, but as N increases, the improvement in precision decreases).
The second question wasn't really clear: I couldn't determine whether you are asking about how to allocate samples when stratifying for two variables (income level and district), or whether income and/or district relate to the DV of interest.
For the first case, choosing n to be proportional to the population size is preferred (though, with very small subgroups, you may need to oversample in order to get reasonable precision). For the second case, it depends on what the variable of interest is. If continuous, then multiple regression (with dummy coded variates for income and district) would work. If categorical, then logistic regression would likely suffice.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size