Guessing Factor in Item Response Theory

Question

I'm currently trying to learn basics (nah who am i joking I'm only trying to learn the concept) of IRT since i have a high stake test (college entrance exam) coming in a few months that used IRT but i don't know the model of what they're using let's just assumed they used what majority of other exam model used.

Question from student perspective : from what i know

• If a student answer a question right they will get 1 point

• If they leave blanks or answer it wrong they will get 0 point

• after calculating 1/0 they calculate the question which one that have less people answer it correctly(harder), and the easier (all people could answer it).

• after that they merge the calculation to the existing score and range the highest to the lowest, and sort them by the score.

  1. What i'd like to know is there a way / some kind of a variable that eliminate guess or merely luck on a test? (I read on Wikipedia there's a 'c' parameters but that's the parameter of the question not the individual people)

  2. Should I answer the question randomly if i don't know the answer, or is it going to negate my overall probability of getting right in the test? (If not then i thought i could randomly pick and hope for the best)

  3. Should I answer the toughest answer because less people will answer it making me beneficial ?

Sorry if this question not related to irt, i try to find my answer downloading bunch of pdf other research paper that have no correlation to my problem.

Any kind of suggestions or explanation, is really helpful.

Answer 1

IRT can be used in many ways and there are many different models.

First, many tests that use IRT are also computerized adaptive tests. In these, you don't have a choice of which questions to answer. You answer each question in turn and you will get different subsequent questions depending on whether you get it right or wrong. This is a good thing because it means you spend more time answering questions that are approximately the right level of difficulty for you.

Second, whether you should guess depends on how the test works. Most IRT tests will use a method that penalizes you for wrong guesses but rewards you for right guesses in such a way that if you are making pure guesses, you won't gain or lose on average. However, if you can eliminate some answer choices as being incorrect, then guessing will help.

Third, if the test is not a CAT and you have a choice of which questions to answer and you don't have time to answer them all, then getting harder questions right will help you more than getting easier questions right. However, usually, answering easier questions is much quicker than answering hard ones. If you have the ability to do really well on the test (which is implied if you can answer the harder questions right) then you will probably have time to answer all the questions. Any gain from skipping around and answering only some questions is probably offset by the additional time it takes to do this and the increased chance of filling in the wrong bubble.