Good enough principle

from Wikipedia, the free encyclopedia

The good enough principle is a principle for formulating hypotheses in statistical tests . It is used, for example, in empirical social research .

Reason for use

With a very large sample, classic hypothesis tests often have the problem that the null hypothesis is very likely to be rejected. The reason is that two populations are very rarely completely alike in order to be able to reject a (punctiform) null hypothesis.

In a practical sense, it is often irrelevant whether, for example, the intelligence quotients of two large study populations differ by only a quarter of the IQ point or not at all.

The previous motivation not to use punctiform null hypotheses , as they would often be rejected, can in principle be extended to the statistical testing of alternative hypotheses : There will hardly be an alternative hypothesis that has the character of a discrete / individual statistical characteristic value, which is determined by a classic significance test would be confirmed exactly .

Therefore, within the framework of the good enough principle, one often defines a null or alternative hypothesis as a so-called range hypothesis, which is considered to be accepted if the characteristic value determined in the empirical experiment falls within its range (an interval as opposed to punctiformity).

The good enough principle, through the acceptance of test result areas instead of the acceptance of just exact test results, thus provides both of the following:

  • not to present scientifically insignificant group differences as significant and thus to overinterpret them.
  • To interpret experimental results as “the null or alternative hypothesis was confirmed” if the result of the experiment falls within a predetermined range around the alternative hypothesis.

literature

  • Bortz, J. & Döring, N. (2006). Research methods and evaluation for human and social scientists (4th edition). Heidelberg: Springer.