highly significantp < 0.05
Both IRR measures yielded highly significant p-values ( p < 0.05), thus rejecting H0 that agreement was purely chance-based.
Both IRR measures yielded highly significant p-values ( p < 0.05), thus rejecting H0 that agreement was purely chance-based.
Rater 3 reached a marginally significant difference between PACS/Qwen texts overall (p: 4.94e-02), but no significant difference for cases labelled correctly by Qwen (p: 1.48e-01).
Conversely, the other two raters found significant or partially significant differences favouring human-generated reports.