Practical Uses of Statistical Power in Business Research Studies
Practical uses of statistical power in business research studies
An important use of power is in the planning of sample sizes prior to gathering data used to evaluate statistical hypotheses. A number of business statistics texts illustrate this use of power (Anderson, Sweeney, & Williams, 1999; Daniel & Terrell, 1995). Again, however, the presentation is somewhat technical and developed through formulas, and is restricted to simple one-sample tests involving a mean or binomial proportion as the relevant parameter.
This approach suggests an emphasis on the concept of power at the expense of the practical usefulness of statistical power.
The use of power to determine appropriate sample size has important limitations in actual business research studies. First, in many studies, constraints based on various practical considerations limit the maximum sample size that can be obtained, possibly making any sample size recommendation based on power calculations unachievable. More important, in many business research studies, this approach to sample size determination is unavailable because most serious research studies involve statistical testing of multiple hypotheses using differing testing methodologies. Different hypotheses might require correlation or regression analyses, the use of structural equation modeling, analysis of variance, and a simple paired t test. Some of these methods (correlation, ttest) permit sample size determination based on power, but others (SEM) do not. Thus, in most research studies, the notion that data are gathered solely to evaluate a single hypothesis is blatantly simplistic. Accordingly, the use of power for sample size determination is not particularly helpful in most such studies.
Therefore, how can power have any practical, beneficial use in serious business research that uses multiple statistical hypothesis testing methodologies? In this article, the author proposes the use of power information as part of the decisionmaking
An important use of power is in the planning of sample sizes prior to gathering data used to evaluate statistical hypotheses. A number of business statistics texts illustrate this use of power (Anderson, Sweeney, & Williams, 1999; Daniel & Terrell, 1995). Again, however, the presentation is somewhat technical and developed through formulas, and is restricted to simple one-sample tests involving a mean or binomial proportion as the relevant parameter.
This approach suggests an emphasis on the concept of power at the expense of the practical usefulness of statistical power.
The use of power to determine appropriate sample size has important limitations in actual business research studies. First, in many studies, constraints based on various practical considerations limit the maximum sample size that can be obtained, possibly making any sample size recommendation based on power calculations unachievable. More important, in many business research studies, this approach to sample size determination is unavailable because most serious research studies involve statistical testing of multiple hypotheses using differing testing methodologies. Different hypotheses might require correlation or regression analyses, the use of structural equation modeling, analysis of variance, and a simple paired t test. Some of these methods (correlation, ttest) permit sample size determination based on power, but others (SEM) do not. Thus, in most research studies, the notion that data are gathered solely to evaluate a single hypothesis is blatantly simplistic. Accordingly, the use of power for sample size determination is not particularly helpful in most such studies.
Therefore, how can power have any practical, beneficial use in serious business research that uses multiple statistical hypothesis testing methodologies? In this article, the author proposes the use of power information as part of the decisionmaking
References: Edward P Markowski;Carol A Markowski; Practical uses of statistical power in business research studies Journal of Education for Business; Washington; Nov/Dec 1999