Abstract

Sampling design for social studies have very specifics requirements, that are generally ignored in the scarce bibliography about matter. The common assumption is that mathematical statistical theory applied in the testing in industrial control quality, pharmacology, medical therapy, etc. is adequate for social populations studies. The problem of the multivariation, the bringing up to date of sampling frame, the variance estimation for sample sizing, in social studies have exigencies no solved in the Sampling Theory. We confront the un-recognized sampling paradoxes: the more used probabilistics samples make use of non aleatory procedures for random selection of units; for estimate standard deviation needed to design a probabilistic sample, you can use a non-probabilistic sample; you adapt the sample size considering only one standard deviation of only one variable; however, you denominate “probabilistic” the whole sample for all other variables and its interrelations. Reviewing the principles of sampling – trying avoid to use mathematical formules- we analyze the troubles for truly probabilistics designs, and, the basis of the validity of nonprobabilistic procedures.