This topic is rather complicated. But there are tools out there that help us do the job. If you just need to know the sample size, just follow the formula strictly.
All we need to consider:
Alpha level = type I error = the level at which you predict that your rejection of the null hypothesis may go wrong. Of course this level should be as low as possible. By definition, it should be at least as low as 0.05. At this level you predict that there is only 5% of chance that your rejection of null hypothesis goes wrong.
Predictors = independent variables = explanatory variables. It can be sex, age, number of children, whatsoever, but you have to make a decision of how many predictors will be there in your sample.
Statistical Power level = by default, it should be .8 or more. I will go back to this issue in a later time.
Anticipated size effect (f2) = choose between 0.02, 0.15, 0.35. Why? I will go back to this later.
