Calculation of Power in Chi-Square and Likelihood Ratio Chi-Square Statistics by a Special Sas Macro

Abstract

The goal of this study was relatively analyzed as to power in Chi-Square and Likelihood Ratio ChiSquare Statistics by using SAS special macro which is presented in Appendix. For the aim, data sets regarding questionnaire responses of 107 refugees were utilized. Contrary to other data sets (had power values with high-level), sample size for only data set 3 having power values with low-moderate level for both statistics were artificially increased from backward to forward and optimum samples sizes for Ch-Square and other were determined as 280 and 170, respectively. As a result, it was concluded that power of Chi-Square and Likelihood Ratio Chi-Square Statistics changed to some factors: the size of sample and combinations of all cells' frequencies of contingency table. Besides, it is possible that researchers can determine sample size which is suitable for each data set by means of special SAS macro in appendix. Moreover, ones should not forget that power concept in any statistic technique means reliability. © 2006 Asian Network for Scientific Information.