In a student well-being dataset (N=148, 7-item 5-point Likert), a polychoric CFA with WLSMV gives ωtotal=0.86 while α=0.73 using R (psych + lavaan). Would you report only ω with a brief note on violated tau-equivalence, or present both and justify the discrepancy — what have reviewers favored in your experience?
I’d report both — lead with ωtotal=0.86 as the reliability from your polychoric CFA/WLSMV, and note in one sentence that the lower α=0.73 stems from violated ‘tau-equivalence’ with ordinal items (psych + lavaan). In reviews I’ve gotten fewer pushbacks when I also bootstrap a CI for ω and park α in a comparability table, especially if the loadings show clear spread.