การศึกษาแบบมอนติคาร์โล : การเปรียบเทียบอำนาจการทดสอบเอฟ เมื่อข้อมูลได้รับการแปลงรูปในรูปแบบแตกต่างกัน ภายใต้ลักษณะการแจกแจงประชากร 3 แบบ / ธีระดา ภิญโญ = Monte Carlo study : the comparison of the power of F - test when applied to different types of data transformations under three different types of parent distribution / Terada Pinyo
The purpose of this study was to investigate the Power of F-test and type I Error Rate computed from Raw-Score and from four different types of data transformations. Those are Square Root Transformation, Reciprocal transformation, Logarithmic Transformation and natural Logarithmic Transformation. The data for this experiment were obtained through simulation using The Monte Carlo technique. A computer program was designed to repeat the experiment 1000 times in each cases. The Comparisons were made for 3 samples where population distribution types are Normal, Uniform and Logistic. The ratio of population variances are 1:1:1 and 1:2:3 Results were as follow : 1.The power of F-test computed from raw scores can control the error rate from the experimental situations the same as the Power of F-Test when transforms in Square Root Transformation and Reciprocal Transformation when a equal to .05 and .01. 2.The power of F-test computed from raw scores has the same high test power as the Power of F-Test when transforms in Square Root Transformation and Reciprocal Transformation. But those tests have the higher test power than the Power of F-Test when transforms in Logarithmic and Natural Logarithmic Transformation when a equal to .05 and .01 and the ratio of population variances are 1:1:1 and 1:2:3 . 3.When the sample size was larger, the power of F-test computed from raw-scores, Square Root Transformation, reciprocal Transformation, Logarithmic transformation and natural Logarithmic transformation can control the error rate better when a equal to .05 and .01.Recommendations :-When the sample size is small or violate the assumption in equal population variance but the symmetry population distribution can be assumed the researcher should use The Power of F-Test when computed from the raw-scores