Sample Size
The larger the sample the better, more accurate the data you get.
- government would not accept any sample smaller than 250 students, balanced for urban/rural, public/separate, and zone: north/south/central/etc.
- but not much you can do about small class/sample size, except
(a) give test to more than one class (secondary teachers)
- need to balance reusing tests three periods in a row against chance of student's in first class tipping off third class on questions...
(b) accumulate stats over a couple of years
- IF you haven't changed the question AT ALL, you can add this year's results to last years to increase sample size
- If you have changed the question, you can compare this year's stats to last year's to see if they have changed in direction you wanted
- have to be cautious because with very small samples, some changes may just be random chance...
(c) Mostly just do it with your 30 students and just be cautious how you interpret these results
- procedures will give you hints at 30 students, proof positive at 400
Procedure works fastest if you have had students use a separate answer sheet
- separate answer sheet means faster marking, so more time for interpretation
- separate answer sheet means more accurate marking too
- believe me, if you're marking "A,B,C,D", turn the page, sooner or later you're going to turn two pages at once and go 3/4 the way through the booklet before you notice that you're now on the last page and have six more answers on the answer key than you have answers to mark...
- separate answer sheet means much faster for counting you have to do for item analysis
UNLESS you're in grades 1-3, in which case have them write on the booklet because they get confused transferring answer to separate page