Using Multi-level Models to Assess Data From an Intelligent Tutoring System

Intelligent tutoring systems yield data with many properties that render it potentially ideal to examine using multi-level models (MLM). Repeated observations with dependencies may be optimally examined using MLM because it can account for deviations from normality. This paper examines the applicability of MLM to data from the intelligent tutoring system Writing-Pal using intraclass correlations. Further analyses were completed to assess the impact of individual differences on daily essay scores along with the differential impact of daily vs. mean attitudinal ratings