Sample Size and Power (B)

If no consideration is given to the concepts of effect size and power, the researcher may end up very much in the dark as to whether (1) a fail to reject decision is attributable to a trivial (or zero) deviation from Ho or is attributable to the test's insensitivity to detect important non-null cases due to a small sample size, or (2) a reject decision is attributable to Ho being false by a nontrivial amount or is attributable to an unimportant non-null case being labeled significant simply because the sample size was so large. In Excerpts 8.14 and 8.15 [not shown here], we see examples of how murky results can be produced when the six-step approach to hypothesis testing is used. In Excerpt 8.14, the researcher tells us, in essence, that the statistically insignificant results may have been caused by insufficient power. In Excerpt 8.15, we see a researcher who obtained statistically significant results but admits, in essence, that the initial findings may have been caused by an overly large sample size making the statistical tests too sensitive.

(From Chapter 8, p. 217)

Copyright © 2003 Schuyler W. Huck
All rights reserved.