LAUSR

The Trigonometric Box-Cox ARMA Trend Seasonal (TBATS) model has been designed to estimate complex cyclical patterns (e.g., weeks within years) in time series data. This paper seeks to evaluate its applicability to educational data, daily school attendance in particular. Attendance rates in four high schools are analyzed over a ten year period using TBATS to illustrate the presence of both weekly and annual patterns in three of the schools and only weekly patterns in the fourth. The model features are explicated and it is shown how the estimation of weekly and annual cycles enhances the description of the data and improves our understanding of how the assessment of endogenous variability contributes to our understanding of daily high school attendance behavior. R script is provided in an appendix.