Sakila,

the sample size depends on:

• the type of effect size indicator you want to compute (e.g., Cohen's d),
• the expected effect size in the population (e.g., d = 1.5) you want to be able to detect,
• the statistical power you want to have (e.g., 1-beta = 0.8) to be able to detect that effect, and
• the Type I error rate your are willing to accept (e.g., alpha = 0.01) in case there is no effect in the population.

For some of the more common statistical models, the software G*Power may be helpful in computing sample size.