A sociologist wishes to assess the association between two ordinal variables, both consisting of 12 categories. The researcher decides to collapse the categories into a dichotomous form (categories 1-5 = "high" versus categories 6-12 = "low") and then produce a bivariate table. Is this a good idea?
A) No. The researcher is left without statistical measures of association: Cramer's V or phi can no longer be used to assess bivariate association.
B) No. Information is thrown away: The researcher must assume that different scores within the "high" or "low" categories are not meaningful.
C) Yes. Measures for ordinal association cannot be used for variables with more than 10 variables.
D) Yes. If the researcher wishes to impose an arbitrary division point on a variable with many categories, it will greatly simplify the analysis.

Question

Quizplus · Accepted Answer

Collapsing categories into a dichotomous form results in loss of information, as it assumes that differences within the "high" or "low" categories are not meaningful, which can lead to a less accurate analysis.

A Sociologist Wishes to Assess the Association Between Two Ordinal