Statistical power as a contributing factor affecting significance among dissertations in the School of Religious Education at Southwestern Baptist Theological Seminar

Problem. The problem of this study was to determine the difference in power of the statistical test between selected dissertations' hypotheses proven statistically significant and those selected dissertations' hypotheses not proven statistically significant in the School of Religious Education at Southwestern Baptist Theological Seminary.;Procedures. The statistical power of selected dissertation hypotheses was measured using the computer program, Statistical Power Analysis for the Behavioral Sciences (1988), developed by Michael Borenstein and Jacob Cohen. The statistical power of hypotheses which met the criterion for being declared significant were compared to the power of hypotheses which failed to be proven significant. A total of 93 hypotheses were included in the study. A t-test for independent samples determined if there was a significant difference according to statistical power between the two groups of hypotheses. Alpha (.05) was partitioned into one-tail to determine if statistical power was significantly higher in the group of hypotheses which met the criterion for being declared statistically significant.;Findings and conclusion. The statistical analysis revealed that there was a significant difference between the two groups of hypotheses. That is, when the statistical power of hypotheses which were proven significant were compared to the group of hypotheses which failed to be proven significant, the two groups were more different than one would expect by chance. Specifically, the statistical power was significantly higher in those dissertations' hypotheses finding statistically significant results than those hypotheses not finding statistically significant results. Therefore, this study demonstrates that statistical power affects statistical significance. Non-significant results should be expected, more often than not, when research is conducted under low statistical power. Consequently, designing adequate statistical power into quantitative research should be a major consideration in inferential statistical testing.