adaptive testing?
It’s not necessarily a reverse-engineering process. The algorithm is well known and anyone within the field of psychometry understands how it works. I have worked and researched for the past year about related topic so implementation is just a breeze. Simply just working with a bunch of data, then calibration. Ronan SAT has a lot of participants especially when taking parts in an Adaptive test so we can mirror that accuracy. As for the accuracy of the model, its literally the same algorithm. I can’t claim that it will be 100% what you expect on test day but its pretty much quite close to +_ 30 pts of difference due to the sheer difference in size. But as I work thru more optimisations algorithm, maybe in the future we will have better accuracy for smaller cohorts ¯\_(ツ)_/¯