Master of Science
The Chi-Square Goodness-of-Fit Statistic is sensitive to the particular partitioning scheme used to divide the hypothesized distribution into cells, viz., number of cells, cell width, and the minimum number of expectancies required per cell. This sensitivity is due to the fact that different partitionings produce different groupings of the random observations and the calculated expected frequencies on which the Chi-Square test statistic is derived.
Gallucci, Lawrence Berry, "A simulation study to determine the optimum number of cells and expectancy content required for the chi-square goodness-of-fit test" (1970). Theses and Dissertations. 3866.