Joint Attendance Response Profiles in PISA 2022: A Discrete D-Vine Analysis of Saudi Arabia

Attendance indicators are commonly analyzed separately or collapsed into a single score, obscuring how behaviors cluster within students. We analyzed four-category reports of lateness, class skipping, and whole-day skipping among Saudi PISA 2022 participants using survey-weighted ordinal margins linked by a discrete D-vine with design-based inference. The method preserves each item’s ordinal structure, separates covariate-related endorsement from residual co-reporting, and yields joint-profile probabilities. The design-weighted table showed that reporting all three behaviors was 8.95% points more common than expected under independence with the same margins. This focal contrast remained substantial across sensitivity analyses, although cell-level lack of fit remained; the D-vine, therefore, complements rather than replaces the weighted table. Simulations showed little bias under the stated response assumptions but appreciable bias with nonignorable item nonresponse. Socioeconomic moderation was weak, whereas parental associations were item specific. Multivariate ordinal modeling provides information unavailable from separate regressions or a single attendance score.