Statistics Glossary → Multiple marginal independence tests (MMI, SPMI)
Multiple marginal independence tests (MMI, SPMI)concept
Methods that use a single overall test to check whether a multiple-response question (where several options can be ticked) is associated with another variable or with another multiple-response question.
In a multiple-response question ('Which social media apps do you use?') a person can tick more than one option, so the total number of responses exceeds the number of people and an ordinary chi-square test violates the assumption of independent observations. In Bilder and Loughin's (2004) test of multiple marginal independence (MMI), a separate 2×J chi-square is computed for each option and their sum X²_S is used as the overall statistic. For the association between two multiple-response questions, Agresti and Liu's (1999) test of simultaneous pairwise marginal independence (SPMI) applies the same logic to every pair of options.
Because the summed chi-squares are not independent of each other, X²_S does not follow a standard chi-square distribution; the p value is obtained with a Bonferroni adjustment (the smallest p multiplied by the number of options or cells), with the bootstrap or with a Rao-Scott correction. The null hypothesis is that the proportion ticking each option is the same at every level of the other variable. If the overall test is significant, the individual option tests show which options differ.
An everyday example
A survey asks 300 people which modes of transport they use (bus, metro, bicycle, car; several may be ticked) and their age group. Running a separate chi-square of age group against each mode would mean four separate tests. The MMI test gives one answer to the question 'do transport choices vary with age?' by assessing all four options together.
In MerQur
It is offered as the 'Overall MMI test' in MR × Categorical Crosstab and the 'Overall SPMI test' in MR × MR Crosstab, with p values computed by Bonferroni adjustment.
Analyses where this concept appears
MR × Categorical CrosstabMR × MR Crosstab