Hi Iram,

You could use a programme such as G*Power to estimate the sample size you require for a given level of alpha (normally .05), estimate of effect size (e.g., is it important to detect a small effect? Or has previous reseach tended to have found large effects?), and desired power (normally >.80). It's freely available from the internet and not too hard work with.

Regarding the effect of population size, it doesn't seem to have a particularly great effect whether it's 1,000 or 100M. Maybe something strange happens if the population size if very small? Interestingly, I'm running a study at present where the population size is around 50, but I'm not concerned about a priori sample size calculation.

N = 150 is probably above average, but using G*Power provides a much better and defensible rationale for choosing sample size. Report your a priori calcauations for sample size in your procedure and, all else being equal, you work wil be viewed more favorably.

Good luck!

Thank you very much!! that was very helpful :)

Andrew Gilbey

Paul F. Franco