Zmed

<< Click to Display Table of Contents >>

Navigation:  Sample studies >

Zmed

Previous pageReturn to chapter overviewNext page

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

 

I will use Excel and Lertap to display results from a screening test used by a Medical School in Czechoslovakia to pass or fail new applicants (people who wanted to be admitted for medical studies).

 

The test had 100 items, all multiple choice.

 

Coefficient alpha was used as the reliability estimate, and it was very good, almost 0.96

 

Alpha reliability estimates above 0.95 are usually found on only the best professionally developed tests. This was a good test!

 

Coefficient alpha is often referred to as a measure of "internal consistency", of how well the items correlate with each other.

 

Computer programs compute alpha using an equation, but the idea is that we pretend to have two parallel tests by splitting a single test into two parts, and then finding the correlation between the two parts.

 

In general, internal consistency estimates of reliability give values which are higher than parallel-forms estimates of reliability.

 

Something to note about this test is that it was used to screen applicants. A pass-fail cutoff score was used and it was found that:

 

54% of the applicants passed (there were 2,470 applicants in total)

 

if they could be tested again, 230 applicants, about 9%, would change classification -- some of the ones who passed the first time would fail, and some who failed the first time would pass

 

on tests of this sort, the "proportion of correct placings" is sometimes used as a reliability figure