Abstract:

This work is concerned with how to measure and, most importantly, how to predict the “amount” of learning in “problem-based” learning programs. Due to difficulties in dealing quantitatively with educational aspects like learning, which are rather abstract when it comes to measuring, this is not a typical problem which can be dealt with easily traditional statistical analysis. The ordinal nature of data collected from surveys requires suitable estimation methods. In this paper we use a method called “estimation by analogy” (EbA), known from its application to software engineering problems, for estimating the amount of learning in a class which is based on historical data containing information from previous learning measurements. The approach is simple since it does not require any mathematical background and assumptions; it is intuitively appealing since it is based on the idea of “finding the most similar cases” and, most importantly, it gives very good results. The method is applied on a dataset from 56 Math/IT classes collected at the end of the fall 2005 from the enrolment of UGRU, United Arab Emirates University.

Keywords:

Estimation by analogy, learning, Validation accuracy.