NCJ Number
74525
Journal
Evaluation Review Volume: 4 Issue: 6 Dated: (December 1980) Pages: 824-842
Date Published
1980
Length
19 pages
Annotation
Relative likelihood analysis, a comparatively unknown form of applied probability theory, is generally a superior alternative to significance testing since it makes possible simultaneous evaluation of the plausibility of families of hypotheses concerning one or more statistical parameters.
Abstract
This article provides a brief introduction to the uses of relative likelihood analysis with current illustrative applications from evaluation research. Most commonly, relative likelihood analysis is concerned with making inferences or testing hypothese about means, proportions, variances, or other statistical parameters, using sample statistics. The data are assumed to be values of random variables distributed in known ways. A likelihood depends upon the probability of an observed sample if a given hypothesis about a parameter value is true. Typically, the hypotheses specify expected patterns of parameter values in different groups involved in the experiment of the survey. The parameter, the model, measures of likelihood, a set of hypotheses plus the data, the a priori standards of what is to be regarded as plausible, and a corresponding set of standard relative likelihoods are the essential elements in likelihood analysis. The problems of measurement of likelihood are discussed at length, and examples are given of such problems. Tabular data, a note, and 10 references are included. (Author abstract modified)