Obraz (126)

/

18. Exolain the followina statistical conceots: a) ranae

b) item analysis    But jt mUSt be

said that without itera analysis, testing - especially multiple-choice testing - is ‘shooting in the dark\ Certainly anyone preparing a test for wide use, for important selection or other purposes, cannot manage without an item analysis. For this to be done properly, there has to be a

preliminary pilot stage when the test is tried out before being put into service. The pilot version should include morę items than you need for the finał version, so that you can throw out any items which prove unsatisfactory while still having enough tried and tested items for the finał version.

Item difficulty

Corning now to actualitem analysis, we should first like to know the difficulty index of each item. This simply means what proportion of the target population gets the item right. It can be expressed as a percentage, giving us a elear idea of the difficulty of the item in comparison with the difficulty of the other items, for that particular target population.

Item discriminatior We may also wish to calculate the discrimination index of each item -

that is, how well it sorts out the better students from the weaker ones This will probably only be important when we reąuire to place the students in a fairly accurate rank order over a fairly wide rangę of ability - for example, in an examination with several grades of Pass and the possibility of a Fail.

There are several ways of calculating the discrimination index of a test item. A simple way begins by ranking all the testees on the entire test, then dividing them into four quartiles- that is, the top 25% of them, the bottom 25% and the two middle 25%s. Then, to find the discrimination index of an item, we see how many of the top ąuartile got it right, and how many of the bottom ąuartile. The ratio between these two numbers is the discrimination index.