CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Multilevel modelling with level-2 missing data: The relationship between student ratings and teacher feelings/practices Authors:  Francesca Bassi - University of Padua (Italy)
Leonardo Grilli - University of Florence (Italy)
Omar Paccagnella - University of Padua (Italy)
Roberta Varriale - Centro Dagum c/o Dip. Economia e Management, Università di Pisa (Italy)
Leonardo Grilli - University of Florence (Italy)
Carla Rampichini - University of Florence (Italy) [presenting]
Abstract: The relationship between student's evaluation of university courses and several characteristics of the student, the course and the teacher is studied. In particular, we exploit data from a survey among professors of the University of Padua collecting information about teacher feelings and practices. Student ratings are nested into teachers, calling for multilevel modelling. However, due to survey non-response, the information about feelings and practices is not available for about half of the teachers, posing a serious issue of missing data at level 2. Note that a standard analysis based on the available observations would discard the entire set of student-level data for teachers with no response to the survey. The problem of missing values in level 2 covariates has received little attention. We focus on Multiple Imputation (MI) methods. Imputation methods for level 2 covariates should assign a value for each level 2 unit to be shared by all nested level 1 units. To this end, a procedure suggested in the literature is to work on two distinct datasets for level 1 and level 2 units, performing separate imputations within each dataset while using the results from one in the other. We consider several imputation strategies, including methods exploiting level 1 information to improve imputation of level 2 missing values.