Soufyane -

I have Thompson's book, 3rd ed, Wiley. 2012, as well as Cochran's Sampling Techniques, and others. I suppose you must be referring to simple random sampling, for continuous data. You need a good guess at the population standard deviation, and a desired standard error.

But beware that there are other methods/approaches for collecting a sample to infer to a population using various estimations (called "predictions" if using regression, as we then have a random variable). Sample size needs depend upon method/approach (various probability-of-selection designs, or model-based approaches, or a combination of these), standard deviation(s), and required standard errors. Some methods result in much more complex 'formulas' than others do.

The kind of data matters. A great deal has long been done for both continuous data and proportions (yes/no data).

Note that some approaches are more "efficient" than others, generally meaning they result in lower variance (standard error) or a smaller sample size requirement for the same standard error goal. Note that bias is generally assumed not to be an issue when developing these sample size requirement 'formulas,' so you should make certain that that is the case. There are different kinds of bias, and sometimes there is some in an equation. There may be some tradeoff. This should be kept in mind. But I expect you are just looking at the simple random sampling case I mentioned above.

If that does not answer your question, you may need to be more specific.

Cheers - Jim

simple random sampling techniques

Simple random sampling