Fit crossed-effects models for reader judgments

Fit a mixed-effects model with crossed random intercepts for readers and stories to determine whether the reported reader-preference effects persist after accounting for the crossed dependence structure.

Background

The reader study is crossed because readers evaluate multiple items and stories are evaluated by multiple readers. The paper’s headline analyses use exact binomial tests and bootstrap intervals that cluster on only one factor at a time.

A crossed mixed-effects analysis is needed to model reader- and story-level variance jointly and to assess whether the reported effects remain under the design’s dependence structure.