Readings in Marketing Information Systems: A New Era in Marketing Research
George B. Saunders, Samuel V. Smith, Richard H. Brien, James E. Stafford · Journal of Marketing · 1970
wrong 100% of the time, even though the experimental design might be such that actual rejections would take place 5% of the time. The book contains a few other vague and misleading definitions and passages. For example, after an illustration involving the rejection of a null hypothesis at the 5% level of significance, the authors imply that the results may be suspect because of the small number of observations. A larger n would have reduced the probability of type 2 error (which is irrelevant, ex post facto, when the null hypothesis was, in fact, rejected). It would also have changed the critical F ratio. But there is no reason to have more confidence in a rejection at the 5% level in an experiment involving a large n than in an experiment involving a small n.