Type I and Type II Errors
Because the decision to reject or fail to reject Ho is fully inferential in nature (being based on sample data), there is always the possibility that a Type I or Type II error will be committed. You must keep this in mind as you read technical research reports, as most researchers do not allude to the possibility of inferential error as they present their results or discuss their findings. In certain cases, the researcher simply presumes that you know that a Type I or Type II error may occur whenever a null hypothesis is tested. In other cases, the researcher unfortunately may have overlooked this possibility in the excitement of seeing that the statistical results were congruent with his or her research hypothesis.
When reading research reports, you will encounter many articles in which the researchers talk as if they have discovered something definitive. The researchers' assertions typically reduce to the claim that "The data confirm our expectations, so now we have proof that our research hypotheses were correct." Resist the temptation to bow down in front of such researchers and accept everything and anything they might say, simply because they have used fancy statistical techniques when analyzing their data. Remember that inferences are always involved whenever (1) confidence intervals are placed around means or differences between means and (2) null hypotheses involving one or two means are evaluated. Nothing is proved by any of these techniques, regardless of how bold the researchers' claims might be.
(From Chapter 10 in the 6th edition, p. 232)
Copyright © 2012
Schuyler W. Huck