Reliability of assessments in engineering education using Cronbach's alpha, KR and split-half methods

An attempt to apply reliability measurements to module assessments in engineering degree programmes is presented in this article. Three techniques for estimating internal consistency - the Cronbach's alpha, KR 21 and split-half methods - were employed in the study. Ten-year data of examination marks were used. The data comprised 13 modules of small- to medium-size classes involving 723 students taught at the university to BSc/BEng engineering degree level. Overall, a majority of modules gave acceptable reliability coefficients of 0.4 to 0.8, based on results obtained from all three methods. A strong correlation was found between Cronbach's alpha and the split-half method. Good correlation of KR 21, with alpha coefficient and split-half methods, occurred only for alpha values exceeding 0.3. It was shown that internal consistency measurement can provide meaningful evaluations of module examinations or assessments in engineering study programmes.