ELEMENTS OF VARIANCE ANALYSIS, EVALUATION OF DIFFICULTIES BY QUESTIONNAIRE

Osmar D. Vera,C. Batanero

Abstract:In this paper we present a study where we assess the difficulties and errors in understanding some elements of Variance Analysis (ANOVA) by a sample of 224 Psychology students, after finishing a Data Analysis course. We analyze the selection of an ANOVA model, the understanding of assumptions, and the associated linear model, the computation of ANOVA in different models, the interpretation of results. These results provide information in an area where no much previous research is available. INTRODUCTION AND BACKGROUND Statistical inference plays a prominent role in various human sciences, including Psychology, with its research based on data collected in population samples, which needs to extend its conclusions. However, the use and interpretation of statistics for publications in Psychology is not always appropriate, as shown in several reviews (Borges, San Luis Sanchez, & Canadas, 2001; Batanero & Diaz, 2006). Also occur in college students, as shown by many studies (Vallecillos 1994), most of which focus on understanding the concept of level of significance, and essential interpretation to determining if the presence of some factor is significant for Variance Analysis. But while, thanks to statistical software, performing the calculations associated with variance analysis is very easy today, teaching concepts and inferential reasoning is much more complex, which explains the many difficulties currently apparent in how inference is applied. In this paper we evaluate how certain basic concepts of variance analysis are employed by a group of Spanish psychology students who, after completing a course on inference, were given a short questionnaire designed to assess how well they understood the following: what determines the selection of a particular ANOVA model; how assumptions regarding the appropriate associated linear model are understood; how ANOVA is calculated in different models, and how results are interpreted. Below we describe the background for our investigation and the sample and method used, discuss the results, and conclude with some reflections on the implications of our research for teaching. Are almost nonexistent research related to learning or teaching variance analysis, its various models, choice of model assumptions and verification thereof or the interpretation and understanding of how a table of variance analysis or experiments design is obtained by which we think that our work provide new results. Rubin and Rosebery (1990) Planned and found a teaching experiment aimed at studying the difficulties of interpretation of some of the basic ideas of experimental design. One is the difference between dependent and independent variable. Another important factor in the variance analysis is the concept of interaction. The presence of interaction is what stands out in ANOVA models on several factors, as it allows to evaluate the effect varies as a factor on the response at different levels of another. Rosnow and Rosenthal (1991, p. 574) indicated that the interaction is the result most universally misunderstood in Psychology. Green (2007) also emphasized that students have difficulty understanding the concept of interaction in variance analysis. In sum, the number of elements involved makes research on ANOVA complex, and few studies have been done that directly monitor what students actually learn or test whether learning objectives have been achieved at the end of the course. In our research project, we, in collaboration with the art teacher, monitored students' progress during a two-year data analysis course and included relevant questions as part of their final evaluation. Below we describe the methodology employed and discuss the results.

Psychology