The impact of a school-based optometry intervention on grade point average (GPA) was analyzed in order to evaluate the program's effectiveness for policymakers. Data were only available for students who received glasses, and included complexities such as inconsistent GPA scales across grade-levels, incomplete follow-up, clustering effects, and differing intervention dates.
Math and reading GPA of 879 students were evaluated retrospectively to assess changes in GPA during 2 years before and 1 year after intervention. Multiple modeling strategies ranging in complexity were employed and visually represented, including piecewise linear and ordinal logistic mixed effects regression. The strengths and limitations of each approach were considered, with respect to its statistical rigor and interpretability.
No single model appropriately addressed all of the statistical nuances of the data structure, and increasingly complex models were less interpretable for the intended audience. The final modelling strategy chosen was a compromise between approaches that sufficiently addressed the statistical complexities of the data and those that yielded suitably interpretable results.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.