Note to myself: workbook saved on Imation USB as: NCERT_RMCS_2017.xlsx

 

Some years ago India's National Centre for Educational Research and Teaching, NCERT, created a mathematics aptitude (ability) test which they wanted to use as a replacement for the SAT, the Scholastic Aptitude Test.

 

I'm not sure why; one reason could be that the SAT was not a free test -- someone had to pay for each test administered, and it may well have been students. I would think that NCERT may have wanted to have a replacement test which was both free and valid.

 

NCERT staff sought to have a measure of "concurrent validity" to see how well their new test, which I have called the "MAT", would correlate with the SAT.

 

I have a copy of the NCERT test validation data collected across several Indian states, sampling over 4,000 students. Each student sat the MAT (94 multiple-choice items) and the SAT (100 multiple choice items).

 

The reliability of each test was estimated by computing coefficient alpha. For MAT, alpha was 0.92, while for SAT it was 0.89.

 

These would be acceptable values were it not for the fame of the SAT, a test which has been used for many many years, and quality controlled. Its alpha figure should have been higher (in my opinion).

 

The correlation between the MAT and SAT scores was found to be 0.76

 

This figure is, in my opinion, low.

 

A review of the SAT item response plots indicated some items did not discriminate. This lowered the alpha value for the SAT and probably the concurrent validity figure too.