This study analyzed the performance in seven statistics units of students who were admitted for a Bachelor of Science (Applied Statistics with Computing) course at University of Eldoret, Kenya in 2012. The units were taught by the same lecturer over a period of three years. The data comprised Continuous Assessment Test (CAT) and exam marks obtained by all the students in this class. The overall mean mark for the seven units was 56.7% ( = 614, = 8.27). There was no significant difference (p = 0.1601) in the mean score of male students value of 56.6% ( = 386, = 8.72) and that one of female students value of 57.3% ( = 209, = 7.50). Results showed that government students performed better than the self-sponsored ones ( = 0.0002). There was a significant difference in mean performance by year ( = 0.0006). The mean score was highest in second year (57.8%) followed by fourth year (56.5%) and lastly third year (53.9%). Simple linear regression models were developed for the seven units with the final exam mark being the dependent variable and the CAT mark being the independent variable. The study established in general there was a positive correlation between CAT marks and final exam marks ( < 0.001, = 0.0441). However only a small proportion of the variation in the exam mark can be explained by the CAT mark. There was significant difference in mean performance from one unit to another ( = 0.0001).