Statistical Analysis for Education and Psychology Researchers

Figure 4.6: A summary of the possible

outcomes of the decision-making

process

4.9 Statistical Inference in Context

My reason for including this last section is to set the classic inferential, what may be
called Fisherian (after the eminent statistician R.A.Fisher) hypothesis testing approach in
the context of modern day use of statistics. The statistics of random sampling, survey and
experimental design, probability theory, formal inference and test statistics, are what I
call here, the Fisherian approach. The importance of this approach as Bartholomew
(1995) notes is that much could be learned from little data provided attention was given
to well designed studies. For this reason alone it is important that researchers should
understand and be familiar with the Fisherian approach to statistical design and analysis
and the debate that surrounds the different schools of inference.

Fisher Neyman-Pearson Debate

Statistical inference has a central role in hypothesis testing and thus far little reference
has been made to different schools of inference. Fisher first put forward the idea that we
can only show that a null hypothesis is likely to be false, we cannot prove that it is true. It
follows from Fisher’s reasoning that the only useful result is a significant result, i.e.,
rejection of the null hypothesis. In Fisher’s view, a non-significant result, that is non-
rejection of the null hypothesis, is of little value because there is not sufficient evidence
to make a definitive decision (to reject the null hypothesis) therefore no decision should
be taken until sufficient evidence (data) is available (to allow rejection of the null
hypothesis). Put simply, Fisher’s case would have been that decisions should be
suspended until the null hypothesis can be rejected.
Neyman and Pearson adopted a more pragmatic approach to hypothesis testing. Their
position was more action-oriented. They claimed that a null hypothesis should either be
rejected or not rejected. The crucial difference between their approach and Fishers’s
position was that non-rejection should not imply suspended judgment and no consequent
action. If the null hypothesis is deemed to be tenable, decisions should be taken as though

Statistical analysis for education and psychology researchers 114