Application to Categorical Error Assessment with Sampling

Robert Gilmore Pontius, Clark University


The contingency table can derive from a sample of the population, particularly for applications to error assessment. If the contingency table derives from a stratified sample where the strata have various sampling intensities, then it is necessary to convert the sample table to an estimated population table for unbiased assessment. This chapter shows how to perform the conversion, then how to interpret the results when the table’s rows are the diagnosed categories and its columns are the reference categories. Relevant software includes the PontiusMatrix42.xlsx spreadsheet available at (Pontius Jr 2020).