Need to know a lot more information to even try to help come up with a sample size.
What are your hypotheses? Are you comparing 2 groups? What kind of variance do you expect in the outcome? Is a t-test the appropriate test to use? What kind of effect size would you want to detect? Are there any background data to help come up with estimates? What kind of statistical power would you want to achieve?
These are the type of questions that one would need to think about in order to help.
Hello Sukaina,
Joe Stanek 's response is correct; you'll need to specify much more about your proposed research in order to arrive at a helpful result. Here's a link that may assist you in thinking about the situation in that type of detail: Article Statistical primer: Sample size and power calculations-why, ...
Good luck with your work.
Sukaina -
It sounds like you would like to know if 30 each would give you a good idea, say of the difference in two means. I suggest confidence intervals over hypothesis tests. Confidence intervals are easier to interpret well. People may often be happy with an arbitrary yes or no result from a "significance test" without fully understanding that result, but confidence intervals provide you with useful information. Perhaps you want to look at something like this, from Penn State's website:
https://online.stat.psu.edu/stat414/node/203/
Note that it is important to see what confidence interval does or does not include zero when you are looking at the difference in two means.
The sample sizes needed to give you an acceptable confidence interval depends upon sigma for each sample population. You estimate sigma, a fixed value, the standard deviation for a population, by using a sample. Standard errors of estimated means are made smaller with larger sample sizes. Standard errors are functions of standard deviation and sample size. They are used in confidence intervals. You mentioned t, which indicates you think the standard form of a confidence interval will do. Other distributions could be used, but not so easily. Sometimes one just looks at an estimate for a "relative standard error."
So, as just stated, "The sample size needed to give you an acceptable confidence interval depends upon sigma for each sample population." After you have an idea of those sigma values, say from preliminary samples, or previous studies perhaps, then you can estimate what sample sizes will give you the "confidence" you want.
Best wishes - Jim
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size