| Sumario: | Three statistical procedures for scalogram analysis are compared and evaluated using sets of real and hypothetical data as well as two distributions of coefficients of reproducibility for 35,000 sets of computer-generated data. The results suggest that Goodman's procedure for computing the standard error of Green's Reps is the most important and useful of the currently available techniques. In view of the work of Green, Goodman, Sagi, Schuessler and others in this area, users of widely distributed families of computer programs should expect any scab gram program to compute "chance reproducibility" and an accurate standard error figure for an observed reproducibility. Investigators using scab gram analysis should provide their readers with this information as well as a definition of error responses, the proportion of non-error responses, marginal frequencies, and the number of cases analyzed.
|
|---|