ABSTRACT
Frequency analysis (see Chapter 5) is a univariate method of identifying a likely population from which a sample was drawn. If the sample data fall near the fitted line that is used as the best estimate of the population, then it is generally safe to use the line to make predictions. However, “nearness to the line” is a subjective assessment, not a systematic statistical test of how well the data correspond to the line. That aspect of a frequency analysis is not objective, and individuals who have different standards as to what constitutes a sufficiently good agreement may be at odds on whether or not to use the fitted line to make predictions. After all, lines for other distributions may provide a degree of fit that appears to be just as good. To eliminate this element of subjectivity in the decision process, it is useful to have a systematic test for assessing the extent to which a set of sample data agree with some assumed population. Vogel (1986) provided a correlation coefficient test for normal, log-normal, and Gumbel distributions.
