In the Categorical platform, the statistical testing options that appear depend on the type of response and the use of grouping variables. Options include tests for multiple responses, response homogeneity, association, relative risk, and agreement.
Test Multiple Response
(Available only for multiple response data with one or more grouping categories.) See Example of the Multiple Response Test. Contains tests for independence of responses across each grouping category:
Count Test, Poisson
Shows or hides a test of independence of rates that uses Poisson regression. The frequency per unit is modeled by the sample categorical variable. The result is a likelihood ratio chi-square test of whether the rate of each individual response differs across grouping levels.
Homogeneity Test, Binomial
Shows or hides the likelihood ratio chi-square test of independence for each individual response level. Each response category has a binomial distribution (selected or not selected).
Exclude Nonresponses
Excludes nonresponses for count and homogeneity tests of multiple response categories. If a row is missing data in all response categories, it is treated as a nonresponse.
Test Response Homogeneity
(Available only for response variables that do not have a multiple response modeling type and when one or more groupings variables are specified.) Shows or hides a report that contains tests for response homogeneity that depend on your grouping variable:
– When the grouping variable is not a multiple response, then one tests for independence of the response across the grouping variable levels. The likelihood ratio and Pearson chi-square tests are provided. See Example of the Test for Response Homogeneity.
– When the grouping variable is a multiple response, then one tests for independence of the response across each of the grouping variable levels. A Rao-Scott Chi-square test is provided.
Cell Chisq
Shows or hides p-values for each cell in the table for a chi-square test of independence. A small p-value indicates a cell with an observed value that is larger or smaller than expected under the assumption that the rows are independent of the columns. The p-values are colored and shaded according to whether the count is larger or smaller than expected. See Example of the Cell Chisq Test.
Compare Each Sample
(Available only for single responses with one or more grouping variables.) Shows or hides a report that contains pairwise likelihood ratio and Pearson chi-square tests for independence of responses across levels of a grouping variable. See Example of Compare Each Sample with Comparison Letters.
Compare Each Cell
(Available only for single and multiple responses with one or more grouping variables.) Shows or hides pairwise likelihood ratio chi-square, Pearson chi-square, and Fisher’s exact tests for independence of each level of the response versus all other levels combined across levels of a grouping variable. See Example of Compare Each Cell with Comparison Letters.
Relative Risk
(Available when the grouping variable has two levels and either the response has two levels or is a multiple response and the Unique occurrences within ID option has been selected.) Shows or hides the relative risks for a two-level grouping variable for each level of the response. See Example of Conditional Association and Relative Risk.
Conditional Association
(Available only when the Unique occurrence within ID option has been selected.) Shows or hides the conditional probability of one response level given a second response level. See Example of Conditional Association and Relative Risk.
Agreement Statistic
(Available only for Rater Agreement responses.) Shows or hides the Kappa coefficient of agreement and the Bowker test of symmetry. See Example of Rater Agreement.
Transition Report
(Available only for Repeated Measures responses.) Shows or hides transition counts and rates matrices for changes in responses across time. See Example of Repeated Measures.
Test Options
Options available in this menu depend on your selected analysis.
ChiSquare Test Choices
Specifies which chi-square tests of homogeneity are calculated for single responses. You can choose between Both LR and Pearson, LR Only, or Pearson Only, where LR refers to likelihood ratio.
Show Warnings
Shows warnings for small sample sizes in chi-square tests.
Order by Significance
Reorders the reports so that the most significant reports are at the top.
Hide Nonsignificant
Suppresses reports that are non-significant.
FDR Adjusted PValues
Shows or hides false discovery rate p-values based on the method of Benjamini and Hochberg (1995).