Father of Chi-Square Tests
  • This calculator provides a test of fit to the tail of a distribution based on a few (up to 75 here) top order statistics assumed to be taken from a much larger random sample. Distributions with Pareto-like tails will not generate significant lack of fit as judged by the chi square test except within the bounds of the error rate for the test specified by the level of significance. Consult the paper by Schuette and Spruill for details. Just enter the data and select the number of cells. The chi square value reported by the calculator should be referred to the upper cutoff of the desired size of a chi-square distribution with df = the chosen number of cells - 1. A test, for example, of size 0.05 based on 4 cells would reject fit of the distribution to a regularly varying tail if the reported chi-square exceeded 7.815.
  • The expected number per cell should be at least 5. So with, for example, 25 observations one should not select more than 5 cells.
  • Numbers may be decimals but with no place markers like commas. Delimiters for the data points entered can be a mixture of spaces and/or crlf's.
  • The data is to be ordered from largest first to smallest last. This can often be done conveniently in a spreadsheet, and then the cells copied, and pasted directly or after an intermediate pasting as unformatted text into a word processor. The test is intended for continuous data and ties are not accepted.