A Comparison of Methods of Analyzing Attribute Data Utilizing Computer Simulation,

摘要

The report describes a simulation study used to evaluate the performance of four methods of analyzing attribute (go-no go) data in a two-factor classification. Sets of proportions were developed within the limitations of seven influencing factors and each set was subjected to analysis by each method. Using the percentage points (alpha's) of the F-distribution which coincide with the tests of significance,the methods were evaluated in light of the influencing factors. It was concluded that the arcsine transformation of proportions method is preferrable for all around use. The zero-one (0 - 1) method has application where the mean proportion defective (P bar) is approximately in the range 0.2to 0.8. (Author)