I recently completed my project at CSU on prediction of students ARMT scores based on 20 questions from the Thinklink assessment. My results produced about a 20% error rate which is much better than the current results provided by the Thinklink testing administration. I’m currently extracting the information provided by the service and will publish some of these predictions soon.
What I have not done is use these predictions as early as November. The predictions I have used before were from February when testing occured in early April. The tests should be equivalent in scaled score difficulty. They may even be the same questions, but I’m unsure of this. Either way, the results will be interesting and help me target specefic student needs.