Quote from EricEnfermero
The prevailing belief mentioned above is somewhat supported by information from the NCSBN, the agency that creates the NCLEX. Check out page 16 of this link: http://www.ncsbn.org/pdfs/2006_Candidate_Bulletin.pdf
It says that, 'Candidates with very high or very low abilities tend to receive minimum length tests.' While you're right that 'total failure' would not be a good characterization of anyone's exam just based on the number of questions, we do know that the computer had much less trouble making a pass/fail decision if an exam cut off at 75 as opposed to 265.
We often (correctly) say here that the number of questions is irrelevant to predicting a pass/fail result, but at the same time we can't say that it's arbitrary and unrelated to performance.
While I agree with this, it also has something to do with consistency, which is also another indication of being highly identified, either way, by ability.
Even if answering questions at a very high level or low level, if your answers are erratic, the test will keep going. It doesn't just measure your ability but the 95% confidence that it can predict that the NEXT question it would ask, at that level, would result in a 50/50 chance of you answering correctly.
For example, if the bar is say, 50%, even though you are answering in a range of 90%, if your 'range' of variability is wide, say 85-95%, the test will keep going until it narrows it to a defined point. It would not matter that the whole range of questions is sufficiently above passing.
This is evident from the fact that almost half of examinees pass if they don't finish the test. In order to do THAT, for each of the last 60 questions, the difficulty level must remain above the passing standard (which doesn't mean you have to answer every question correctly). It does mean that the entire range of the last 60 questions was above the passing standard.
A rasch model (of which, the NCLEX is one) establishes the predictability of a particular point. Until it accurately measures that point in an individual, it doesn't presume that the range of variability is set.
Only if 265 questions are answered (or time expires) does the computer stop looking for statistical accuracy and actually look only at the level of difficulty of the examinee.
So, yes, a high or low level of ability would tend to decrease the variability of an individual's answers and narrow to the defined point of that individual's performance in a minimal length of questions. That creates a 'tendency' of high and low performers to have a minimal number of questions. But, that is not an absolute.
Just because the test DOESN'T stop at 75 is no real indication of your level of performance. It doesn't do test takers any good to 'freak' because at question 76, they must be 'close to failing', or borderline. In fact, because the test is looking for statistical accuracy, the number of questions is UNRELATED to the general level of difficulty at any given question.
In other words, the test doesn't 'keep going' because you are borderline in your scoring. It keeps going only until it achieves a 95% statistical probability that it knows your level. At that point, your level is determined and the test stops.
As such, the number of questions really IS unrelated to how you are doing on the test.
Does that make sense?