The effect of sample size on the statistical test power

Abstract

This study aimed to reveal the effect of size of the sample and the number of independent variables on the strength of the statistical test in some of the statistical methods commonly used in educational studies. To achieve the aims of the study, the researcher used the descriptive approach, and the study tool was represented by a table consisting of a group of rows representing individuals and a group of columns representing variables. The study sample included a random sample of statistical vocabulary that ranged between (10) items and (330) items, depending on the nature of the statistical test used. Through the results it was concluded that with increasing sample size, noted that increasing of test (T) strength for one sample, when the sample size was (n=30) the test power was (0.877) while the value of (Beta) declined to (0.12), as the result of the test (T) for two interrelated samples the size of the sample (n=30) was very appropriate for high- power test (0852) and the value of (Beta) declined to (0.148), in the test (T) for two independent samples the power of the test was high (0.773 – 0.856) and the value of (Beta) is low (0.228- 0.144) when the sample size between 40 – 50. After reaching these results, some recommendations has been proposed including increasing the size of the sample to the number that gives the strength to the suitable test estimated at about (0.8) in the field of educational sciences, and decrease the value of the error of the second type which is estimated at about (0.2) in the field of educational sciences.