Reliability
There are several different "kinds" of reliability. One kind (test-retest) looks at whether someone gets the same score if they take the test at different points in time. Another kind (internal) refers to the extent parts of the test "hang together." For example, if we split your test of Knowledge of Basket Weaving into two halves, would scores on the first half pretty much mirror scores on the second half?
Validity
Similar to reliability, there are several "kinds" of validity when it comes to tests. Without boring you to tears, validity essentially refers to the extent to which the results of the test can be interpreted the way you want. If you're looking for someone's math skill, is that what the test is measuring? That's the most common layperson definition of "validity", although there are the traditional concepts of content validity, criterion-related validity, construct validity, and face validity.
Having fun yet? Two final points: you can't have validity without reliability. You can't have a test that measures what you want it to measure if someone gets a wildly different score every time they take it. And the corollary: just because a test is reliable doesn't mean it's valid (more on this in a sec).
The fun part of the post
Okay, so we've done the requisite introductions. The real part of this post has to do with a company called Despair, Inc. If you haven't heard of them, Despair sells "de-motivational" products that mock the traditional motivational posters you've no doubt seen in offices everywhere.
The company recently introduced some new products and one of them is related to my discussion above above. Take a look:
Although I suspect this relates more to individual performance, the point is the same when applied to tests--a consistent (reliable) test is good only if it's measuring what you intend it to.
I may need to get this one, although I already have three of their other posters hanging in my office. I use them to gauge sense of humor:
No comments:
Post a Comment