6533b832fe1ef96bd129a2ed

RESEARCH PRODUCT

Classification trees for multivariate ordinal response: an application to Student Evaluation Teaching

Antonella PlaiaVincenza CapursiMariangela Sciandra

subject

Statistics and ProbabilityOrdinal dataMultivariate statisticsComputer sciencebusiness.industryOrdinal ScaleDecision treeGeneral Social SciencesDecision tree Ordinal response Student Evaluation of Teaching Distances02 engineering and technologyMachine learningcomputer.software_genre01 natural sciencesOrdinal regression010104 statistics & probabilityStatistics0202 electrical engineering electronic engineering information engineeringProfiling (information science)020201 artificial intelligence & image processingTree (set theory)Artificial intelligence0101 mathematicsbusinesscomputerOrdinal response

description

Data from multiple items on an ordinal scale are commonly collected when qualitative variables, such as feelings, attitudes and many other behavioral and health-related variables are observed. In this paper we introduce a method to derive a distance-based tree for multivariate ordinal response that allows, when subject-specific characteristics are available, to derive common profiles for respondents giving the same/similar multivariate ratings. Special attention will be paid to the performance comparison in terms of AUC, for three different distances used as splitting criteria. Simulated data an a dataset from a Student Evaluation of Teaching survey will be used as illustrative examples. The latter will be used to show the performance of the procedure in profiling students by identifying which features of their experience are most closely related to their expressed satisfaction.

yearjournalcountryeditionlanguage
2016-10-06
10.1007/s11135-016-0430-2http://hdl.handle.net/10447/203068