Abstract:

The objective of this paper is to present the application of collaborative and non-collaborative activities to evaluate students in a statistics course of a computer science engineering degree. The face-to-face course is compulsory and includes basic statistical topics. It is taught in the second semester of the degree syllabus. The course contents are descriptive statistics, probability distributions, an introduction to inference, analysis of variance, and correlation and simple regression models. Emphasis is placed on the instrumental use of methods. In relation with simple regression models, the syllabus includes the basic model assumptions, parameter interpretation, residuals analysis, inference on the model and its coefficients, and model evaluation. In the 2014_15 academic year, this part of the course was assessed in the classroom, in the computer laboratory session and with an exam at the end of the semester. The evaluation in the classroom consisted in a short problem, which involved the application of the formulas to estimate the intercept, and slope of the model, and a 95% prediction interval. Each student presented a written answer, that they could find using the course teaching materials, their notes, bibliography recommended for the course or any other information source. The activity could be collaborative or non-collaborative, with a consensus or a non-consensus approach. In the computer laboratory practice students worked in teams of two or three people. They used statistical software to apply descriptive tools (sample covariance matrix, correlation, scatterplot), and a simple regression model. A small sample data set (25 bidimensional numbers), that posed several questions on a computer system, was analyzed. At the end of the session, each team had to present a written consensus answers to an evaluation task that involved another example. Using the software output, students were asked to write and estimate a model that established the dependence of height on age of a sample of 263 school children, with the techniques taught in the lectures and in the practice. The intercept and slope identification in the output, their interpretation and significance tests, and the conclusions of the model analysis of variance, were another assessment questions. They were asked to indicate the unexplained variance magnitude, from the parameters in the output that quantified it. The application of the regression equation was evaluated by the estimation of a forecast error and a prediction interval. In the 2014_2015 academic year, the exam had a problem with software output on the application of a simple regression model to study the performance of a vector ordering algorithm. The problem implied the use of some of the tools that were evaluated in the computer laboratory session. In this work, this year evaluations results are presented for a sample of 79 students of two different groups. The average mark obtained in the exam simple regression question, is worse than in the classroom and computer laboratory tasks in both groups. In one group there is a linear correlation between the average marks in the exam and the computer laboratory task. In the other group the correlation is significant between the classroom activity and the exam problem.

Keywords:

Assesment, statistics education, collaborative evaluations, simple regression models.