I would like to conduct research about the effect of sleep on frailty in smoker older people by dividing the participants into 3 groups (no smoker, formal smoker, and current smoke). How can I calculate the sample size in case I know the population?
If you have many segments, Stratified sampling technique is the ideal one, if you got to know the population size. For example, there are 300, 200, and 500 numbers in your population of 1st, 2nd and 3rd groups respectively and you aim to target to have 100 samples in your study. Hence, your sample selection should be 30, 20, and 50 respectively from each group. This is because your population is 1000 in number and your sample size 100 is the 10% of 1000. Accordingly, the 10% of each group can be selected from each group. Also note that your study should count a minimum of 5% of the population in your sample size.
However, if your population is unknown, then you have to select the samples as per your study requirement. In this context, I do agree with Tadesse Alemu.
Hope this explanation can help you understanding about your sampling requirement.
Not sure I understand your question. Taken literally, to calculate the number in each condition you count how many are in each condition. If you mean how many to recruit for a future study, that obviously depends on the design as Tadesse Alemu
, population characteristics like Samithamby Senthilnathan says, and most importantly why you are doing the study. Because I would think anyone would realize these are necessary (this seems more obvious than not knowing to count), I will taken your question literally. You count the number in each condition.
Depends on the analysis (ex. t-tests, ANOVA, Bonferroni, two-sample binomial, etc.). What is your analysis? A generic suggestions is plan on doing a t-test. Write the mean and sd for each group (or make an educated guess). The do the t-test sample size calculation for all three possible comparisons: A vs. B, A vs. C, B vs. C. Whichever calculation produces the LARGEST group sample size, is what you should use for ALL groups. This is a generic suggestion since I don't know the specifics of the sampling design, data types and analysis to be of more specific.