Skip to content
Omnikalc

Kuder-Richardson KR-20 and KR-21 calculator

Enter the number of items, the sum of p × q over the items, the mean and the variance of the total scores to get KR-20, KR-21, the standard error of measurement and the number of items for a target reliability.

Updated on
Your inputs

Value out of range

Value out of range

Value out of range

Value out of range

Value out of range

Results

KR-20

0.852

KR-21 (assumes equal item difficulty)

0.806

Items needed for the target reliability

32

Standard error of measurement (from KR-20)
1.74
Standard deviation of the total scores
4.53
Mean item difficulty (proportion correct)
0.600

The Kuder-Richardson KR-20 and KR-21 calculator gives the reliability of a test scored right or wrong from the number of items, the sum of p × q, and the mean and variance of the total scores, with the standard error of measurement and the number of items needed for a target reliability.

How the calculation works

  • Enter the number of items and the sum of p × q over the items.
  • Enter the mean and the variance of the total scores, and a target reliability.
  • Read KR-20, KR-21, the standard error of measurement and the items needed.

Example

A test of 20 items with a sum of p × q of 3.9, a mean of 12 and a variance of 20.5 has a KR-20 of 0.852 and a KR-21 of 0.806. The standard deviation is 4.53 and the standard error of measurement 1.74. The mean item difficulty is 0.600. To reach a reliability of 0.9, the test needs 32 items.

Why it matters

The calculator helps teachers, test developers, psychometrics students and researchers who evaluate a test of multiple-choice or true-false items, check that it is reliable enough for its use, and decide how many items to write for a new version of a test.

Good to know

Compute p for each item as the proportion of examinees who answered correctly, and sum the products p × q. Use the variance of the total scores with the same divisor as in your software, and the same group of examinees for all the figures. KR-20 assumes that the items measure a single trait and are scored 0 or 1. KR-21 is a lower bound that is convenient when only the summary statistics are known. A high value does not by itself prove that the test is valid.

The Cronbach alpha calculator handles items with several score levels, the Spearman-Brown calculator predicts the effect of test length, the item difficulty and discrimination calculator analyzes items and the reliable change index calculator uses the reliability for change scores.

Frequently asked questions

What is KR-20?

The Kuder-Richardson formula 20 is an estimate of the internal consistency of a test whose items are scored right or wrong. It is the special case of Cronbach’s alpha for dichotomous items. It is KR-20 = k / (k − 1) × (1 − Σpq / σ²), where p is the proportion correct on an item, q = 1 − p, and σ² is the variance of the total scores.

What is the difference with KR-21?

KR-21 simplifies KR-20 by assuming that all the items have the same difficulty, so that Σpq can be replaced by M × (k − M) / k, using the mean score M. It needs only the mean and the variance and is easy to compute, but it is never higher than KR-20 and underestimates the reliability when the difficulties vary.

Which value is acceptable?

For classroom tests a reliability of 0.70 or more is often considered acceptable, 0.80 is good, and 0.90 or more is wanted for high-stakes decisions about an individual. A short test, or a test of items with very different content, tends to give a lower value. The reliability also depends on the group tested.

How can reliability be raised?

Adding items of similar quality raises it, as predicted by the Spearman-Brown formula. The number of items needed to reach a target is k × r_target × (1 − r) / (r × (1 − r_target)). Removing items that do not discriminate also raises it, and so does making items of medium difficulty, around 0.5 to 0.8.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.