non linear testing brainstorm

abstract nonlinear testing, et data from wrong answers, it the person guessing? did they make a common mistake? what was that mistake and what are the most common mistakes? break apart what each question is asking the greatest difficulty will be recognizin all the different kinds of mistakes. we want to identify gaps in knowledge rather than successes. more difficult questions with lower full success rate but encourage partial credit multiple choice make more questions with multiple parts but keep knowledge areas more discrete (not having answers build on one another) aptitude tests that also help to diagnoise psychological problems multiple choice math should be word bank instead?