Statistical power analysis: An application in social sciences

The power analysis is a method that enables to plan a highly valid and reliable research and to guarantee the validity, reliability and sensibility of the results of a research. Statistical power, in addition, is an approach which is used to evaluate to what degree the decisions obtained as a result of statistical tests used to test an aim are valid and reliable in terms of probability values. The power analysis is different from the other statistical methods to a great extent. Various statistical analyses begin the analysis with the existing data and focuses on the comments of the consequences. Nevertheless, power analysis is meaningful before data collection and significant in the process of setting the hypothesis.In this context, the major purposes of this thesis are;i.To contribute researches in social sciences by creating awareness about what the power analysis means.ii.To guide researchers about how the power analyses can be employed in a social science study.In this study, the master and PhD dissertations which are successfully completed between 2005-2012 years in ?Human Resources? area and registered to the Council of Higher Education (YÖK) National Thesis Center are investigated to detect statistical power levels. This study is expected to contribute to the existing statistical power knowledge significantly, since there is no exist other study similar to this research.In the literature part of thesis detailed information has given about statistical power and also in the application part statistical power analysis process on the an example has given. Cohen?s power tables and G-Power 3.1.3 statistical power analysis program, the statistical powers of the masters and PhD dissertations included in the study are found as 0.23, 0.72 and 0.92 for small, medium and large effect sizes respectively. These power values are less than 80% level which is suggested by Cohen to be able to determine small and medium levels effect sizes except large effects. On the other hand, the power values attained as a result of this application are relatively greater than those of the existing literature. Moreover, in the application part, the outcomes of the significance tests employed in the dissertations are examined and 1685 of the tests (56.73%) are found to be insignificant and 1285 (43.27%) are found to be significant out of 2970 total significance tests. The reason behind the failure of the majority of the significance tests may have been associated with low statistical power. Thus, the significance tests which result in insignificant results will be better if they are repeated after adequate statistical power is assured.