Cronbach's α and KR-20

Question 1

What does Cronbach’s α measure?

Answer 1

Cronbach’s α measures the internal reliability and consistency of a test.

The possible values range from 0 (unreliable) to 1 (maximum reliability)…but an α of 1 can indicate that the items are too similar to each other, since there’s no variation.

Question 2

What are good and bad Cronbach’s α scores?

Answer 2

Over 0.9: very good
0.8-0.9: good to very good
0.7-0.8: sufficient to good
Under 0.7: bad to questionable

Question 3

Cronbach’s α measures reliability. How does this relate to checking the validity?

Answer 3

The test must be checked beforehand for content validity. Cronbach’s α measures the reliability in terms of repetition.

If a test is not reliable, it is not possible that it is valid in terms of content.

Question 4

Case study
α is lower than 0.6. What do you recommend?

Answer 4

This means the reliability of the test is too low to base decisions on. Consider the test result formative only.

Question 5

Case study
α is between 0.6 and 0.8.

Answer 5

The test has a low reliability. If it can be compensated with other tests (covering the same learning goal), then OK.

If the test cannot be compensated, remove the questions with low or negative Rit, as long as this doesn’t affect representativeness.

Question 6

What is KR-20? When is it used?

Answer 6

The KR-20 (Kuder-Richardson Formula 20) coefficient is a measure of test reliability, similar to α.

KR-20 is specifically for dichotomous items. It assesses the consistency of responses across test items.

Question 7

What is a good KR-20 value?

Answer 7

The ideal value is (well) over 0.7.

Question 8

How many questions does a test need to have for a reliable α or KR-20 value?

Answer 8

40-60 questions minimum gives a good reliability value (more questions, more reliable reliability value).