Figure B1

The meter separates genre pairs only where a real difference exists

Appeal difference in score points for five genre pairs across four stations, each with its 95% confidence interval. Three pairs where the genres should genuinely differ land clearly off zero. Two pairs built as negative controls, where no real difference is expected, both cross zero: the meter declines to separate them.

Pairs expected to differ; intended null pairs. Source: five genre pairs across four stations, null pairs included; frozen values from the published control.