Methods: Using National YTS data, probability of student non-response will be modeled as a function of school characteristics, student demographics, whether the student took the NYTS as a makeup, and self-reported 30 day absenteeism. Propensity models will help identify the best predictors of non-response. We will assess whether prevalence rates are significantly different for students taking make-ups and as a function of higher 30-day absenteeism rates. Item analysis will consider item type and position to test the hypotheses that item response decays as the questionnaire nears the end, and increases as grade level increases.

Results: Higher prevalence rates among students with higher levels of absenteeism or who completed the YTS as a makeup suggest ways to factor such data into weighting adjustments to mitigate non-response biases. The lack of correspondence between grade and item response rates, particularly later items, suggests that fatigue and lower reading abilities do impair YTS completion by younger students.

Conclusions:

Absenteeism data, school-level and student-level demographics, and whether students took the YTS as a makeup, can be useful in conducting non-response adjustments. Item non-response analyses will identify items with high missing data rates, perhaps arising from perceptions of the saliency of the question or lack of understanding.