An Estimation of Short- and Long-Term Error Dependencies in Daily High School Attendance Trajectories

• In Event: 49.061 - Advanced Modeling Strategies for Longitudinal, Multilevel, and/or Survey Data

Abstract

Since nonlinear time series analysis is rarely used in education, this presentation aims to show how the technique can be used to address important questions in that context. The presentation focuses on the use of the Autoregressive Fractional Integrated Moving Average (ARFIMA) approach, which is designed to analyze long-term error dependencies in time series data. Daily high school attendance rates in two schools over a seven-year period are analyzed to illustrate what can be learned about daily high school attendance when ARFIMA is used, and what is concealed by conventional measures of central tendency and variability. The analysis reveals a strong weekly cyclical pattern in both schools, as well as long-term error associations in one of the two schools.

Author

• Matthijs Koopmans, Mercy College