- Thread starter neha_sharma
- Start date

It completely depends on the analysis in question, the desired effect size, and margin of error.

Edit: Also I believe to conduct a power-analysis in SPSS you need to download (maybe buy?) a bolt-on package.

G*Power is a free software system which can be used to predict sample sizes needed to detect differences based on specified inputs.

http://www.gpower.hhu.de/en.html

But as Evelyn13 said, the answer depends on the analysis. If what you are measuring has only a very small difference in the population means, then your sample size has to be large in order to detect it. If the difference in population means is massive, then your sample needn't be very large (assuming fixed alpha and beta). To estimate an effect size you either need a pilot study or refer to previous research (again all depending on what it is you're doing).