Estimating the difficulty of A'level examination subjects in Uganda

Abstract

In order to gain access to institutions of higher learning in Uganda, including universities, all students sit a national examination at the end of A’Level, the scores of which determine their selection into various institutions of higher learning, including university. For most university degree programmes, entry is determined based on the A’Level scores irrespective of subject, essentially implying that the same scores in the different subjects are comparable. In order to investigate this comparability, a generalised partial credit item response model was fit to the A’Level examination results data for the years 2009 and 2010. Science and non-science subjects were hypothesised to load on two separate dimensions of the latent ability scale, and subject difficulty and discrimination parameters were estimated. It was found that science subjects were relatively more difficult than humanities and language subjects, and that they also provided the largest amount of information, although this was for the higher end of the ability scale. Some other subjects like Art and Kiswahili were not only relatively easier, they also provided very little information on the ability scale underlying the other subjects. These findings bring into question the comparability of scores in the different subjects at A’Level, and if student ability based on examination performance can be better represented by integrating information on difficulty levels.