An Introduction to Statistics in Linguistics (6): How to Handle the Number of Occurrences

Abstract

This paper explains how the number of occurrences of forms in linguistic research can be handled and evaluated, dividing the cases into those based on sample surveys from a larger population and those based on complete surveys; trials of numerical calculations are included. There are numerous effective tests for evaluating the number of occurrences in a sample survey. In linguistic surveys, the chi-squared test is often used in the analysis of usage surveys and questionnaires. This test is particularly effective when describing relationships between variables; however, it is important to note that generating accurate results when the expected value is extremely small or when the sample size is extremely large is difficult because of the test’s strong sensitivity to the sample size. In contrast, in the evaluation of the number of occurrences in a complete survey, the frequencies with respect to the entire population surveyed are important. For inter-source comparisons of classical words, it is necessary to use relative frequencies based on the frequencies in the entire population, rather than raw frequencies.