|
Reliability
issues within each category of evidence
An investigation of the reliability of course grades,
or any other quantitative measure of student learning, might entail
the following:
The computation of an alpha coefficient
when the grades are thought to be measuring a single attribute.
Correlations between two different administrations of a test
that determined the grade
-or between even and odd items on the test
-or between the first and second half of the test
-or a correlation between equivalent versions of the test
-or the stability of the mean grades and standard deviations
across several administrations of the test to comparable groups
-or published reliability statistics from test manuals |
Along the same lines, the faculty might explore
the reliability of their grades through correlations of the grades
from each half of the transcript for a random sample of students,
or through the correlations between grades in the same course in
two semesters from a sample of professors, or by examining whether
the variance in the distribution of a faculty member’s grades
(0-4), or the variance in the average grade in selected courses,
is contained within one point or a letter grade. In general, correlations
about .80 yield confidence that the measure is trustworthy and dependable.
|
