Split Half Reliability (Spearman-Brown)

Split-half reliability estimates a test's internal consistency by correlating scores on two halves of the items and stepping the correlation up to full length with the Spearman–Brown formula.

반분 신뢰도는 문항을 두 반쪽으로 나눠 반쪽 점수 간 상관을 구한 뒤 Spearman-Brown 공식으로 전체 길이의 신뢰도로 환산해 내적 일관성을 추정합니다.

How reliable is my 20-item knowledge test if I split it into odd and even items?

20문항 지식 시험을 홀수·짝수 문항으로 나누면 신뢰도가 얼마나 나오나요?

The items are divided into odd- and even-numbered halves, the correlation between half scores is computed, and the Spearman–Brown formula 2r/(1 + r) projects it to the full test length. Guttman's split-half coefficient does not assume the halves have equal variances and is reported alongside.

When to use Split Half Reliability (Spearman-Brown)

• A single administration of a multi-item test
• A quick reliability estimate that is easy to explain
• Comparing with older reports that used split-half reliability

When to use something else

• Items are binary and a split-free estimate is wanted, where KR-20 is standard
• Any split should be avoided, where Cronbach's alpha averages over all splits

Assumptions and requirements

• The two halves measure the same construct (parallel halves)
• Items are scored in the same direction
• Subjects respond independently

What the analysis reports

• Correlation between half scores
• Spearman–Brown corrected reliability with 95% CI
• Guttman split-half coefficient
• Cronbach's alpha for comparison
• Item assignment to halves

How to report it

Odd–even split-half reliability was [value] after the Spearman–Brown correction (half-score r = [value]); Guttman's split-half coefficient was [value] (N = [value]).

Related methods

• Cronbach's Alpha Reliability Analysis
• Kuder-Richardson 20 (KR-20)

Run Split Half Reliability (Spearman-Brown) on Quorant