Wallace Judd

Reliability -- Three Types

I've always thought of reliability as a unary concept... something you tried to measure with increasing precision.

Recently I've been reading Guilford, "Fundamental statistics in psychology and education", pp. 449 ff.

Interestingly, he refers to two different types of test: Homogeneous & heterogeneous. He says, "We expect homogeneous tests shall be internally consistent" .. consequently an internal consistency reliability measure is appropriate. Something like a KR-20 or Cronbach's alpha.

But he goes on to say, "The only meaningful estimate of reliability for a heterogeneous test is of the retest variety." He terms this a 'stability' measure, as opposed to the above-mentioned 'internal consistency' measures.

After this enlightening discussion, Guilford goes on to detail a number of internal consistency measures, but does little to elaborate about stability measures.

Since performance tests are largely heterogeneous, it seems we should be pursuing measures of stability.

Can anyone recommend references on stability measures of reliability?

The three measures of reliability Guilford (p.445) notes are:

1) Internal consistency reliability
2) Alternate-forms reliability
3) Test-retest reliability

He says Cronbach proposed names for 2) and 3):
2) Coefficient of equivalence
3) Coefficient of stability;
and he (Guilford) proposed
1) Coefficient of consistency.

I think if in the future we frame our reliability discussions in these terms they will be much more productive.

And perhaps we can stop searching for a single reliability measure that would work for both performance tests and unidimensional multiple-choice tests.
Feb. 13 2009


Post Comment

  For security reason enter the characters shown in the picture below.
The characters are not case-sensitive.

security img
*Word in image