Distortion of Type I Error Rate in the Analysis of Variance for Non-Normal Data
MetadataShow full item record
This study is concerned with the comparison of probabilities of type I errors of various error distributions in the one-way analysis of variance. The probability of a type I error using the F-statisticas obtained for various forms of error distributions and compared to determine if there were any discernible trends. A second objective was to test a method of analyzing the non-normal data, hopefully giving probabilities of type I errors similar to that obtained from' normally distributed data. This method requires the substitution of standard normal order scores (Harter,1961), each multiplied b,y the estimated standard deviation of the original data, for the ordered residuals of the original data. It was hoped that this procedure would in some way "normalize" the data.