Statistical Issues on Multiple Choice Tests in Engineering Assessment

This work studies the psychometric properties of the multiple choice tests used to assess skills in Statistics in the Integrated Masters Degree in Engineering and Industrial Management of the University of Minho. The results show the importance of measuring the quality of the instruments to a fair assessment of the students. Moreover, in the Bologna Process context, multiple choice tests in combination with other learning assessment methods are an alternative for the practical implementation of a continuous assessment. The Classical Test Theory approach and Item Response Models were applied to the data obtained in the tests administered to students enrolled in the academic years 2006/2007 and 2007/2008. The statistics considered are: item discrimination and difficulty, reliability (internal consistency), information function and standard error of measurement. With the aim of improving the items that compose the instrument, we present the poor items and analyse them in terms of specific objectives of learning.