Steve Bunk's 5-PRIME on the null hypothesis was a good, brief piece.1 However, he made the common error of confusing the null hypothesis (the hypothesis against which the research result is tested) with the nil hypothesis (definition: if samples come from populations with identical parameters, the hypothesized difference is zero). The null hypothesis--that is, the one "sought to be nullified"--may be nil. But it also may be a population parameter, or a specified difference between population parameters, against which any difference between the samples are to be tested.
Statistical significance is a function of sample size. With large enough samples, practically every difference will be statistically significant; with small samples, important results may be overlooked if statistical significance is relied on exclusively. Bunk's reminder that statistical significance does not necessarily indicate substantive significance is, for a few reasons, on target.
James P. Shaver
Hyrum, UT
shaver@cc.usu.edu
