Why does psychology mainly rely on the use of inferential statisticsas a measure for drawing conclusions about behaviour? Can the normal distribution be a useful concept for deriving experimental conclusions about the behaviour of individuals?

Any suggestions welcome, thanks, Carole...

-- carole forde (caroleforde@hotmail.com), May 09, 2001

Answers

Contrary to a common misapprehension, most inferential statistics do *not* require that the *data* be distributed normally, but rather, only that the *means* of the corresponding sampling distribution are. That the distribution of means converges on to the normal curve, regardless of the shape of the underlying data distribution, is virtually assured by the central limit theorem (given a large enough sample size), which is not a merely empirical assumption but a deducive mathematical conclusion.

-- Christopher Green (christo@yorku.ca), May 09, 2001.

Hi Carole, your question shook lose a few maybe not so random thoughts in my head, often those experimental conclusions under consideration here are not intended to draw any conclusions at all about the behavior of individuals. Often the researcher is hoping that by lumping the individual differences together they can posit a general law or probablistic statement about us as a group. The business about making statements, it turns out is kind of round-about. For instance, say a percentage of people in Friend, Nebraska have shown to suffer from some disease; OK, what can you say for certain about the next one you see? Not much it turns out, all you can say "for certain" is, the statement: That person has the disease! has X% probablity of being true. This inferential distance between a person like yourself who wants to know specific things about individuals and psychological research might seem frustrating, and yet the impact on individuals can be important if you consider a human factors experiment which helps design a better fire alarm. Best, David

-- david clark (doclark@yorku.ca), May 10, 2001.