Google

In correlational research, when continuous measures produce highly skewed distributions—meaning the data is heavily clustered at one end of the scale—researchers may choose to transform them into dichotomous (binary) variables. For instance, a frequency measure of a rare behavior might be converted into a simple $$0$$ (did not occur) and $$1$$ (did occur) format to facilitate more accurate statistical analysis.

Dichotomizing Skewed Variables

Jouriles and colleagues conducted a complex correlational study assessing the relationships among adolescents' experiences of physical aggression, psychological aggression, and psychological distress. To handle highly skewed distributions in their aggression measures, they first transformed them into dichotomous (binary) variables ($$0$$ for did not occur, $$1$$ for did occur) before computing the correlations among the constructs. They ultimately found that experiencing physical aggression was moderately correlated with psychological aggression, which in turn was related to symptoms of psychological distress.

Example of Complex Correlational Research: Relationship Aggression

In correlational research, how might researchers transform a continuous measure if it produces a highly skewed distribution where data is heavily clustered at one end of the scale?

Learn Before

Related