Agreed and point well made. What I wanted to know was how large a sammple would be needed to give assurance of legitimacy. In addition, as was pointed out, one needs to define demographic groups and do a multivaritae analysis. What I mean here is to put all the information “in the pot” so to speak and see what matters. For example, if one studies any observation and compares it with one element of a demographic group, the answer may not be correct. For example, if one studies an event related to “maleness”, you need also to include all (or as many as possible) of the variables that maleness entails. Examples would be age, weight, race, conditioning etc. One may find out that an observation associated with maleness is really associated with age. Am I being clear? Sometimes I am not as clear as I would want to be.
“What I wanted to know was how large a sammple would be needed to give assurance of legitimacy.”
Actually, the sample number can be fairly small and still give accurate predicitons. For instance you can make a fairly accurate national poll with only a few thousand samples. The problem is getting a sample that is representative of the population and not skewed. A representative sample is much more important than a large sample size.