MerQuR+Statistics Glossary

Statistics Glossary → Spearman-Brown formula (single and average measures)

Spearman-Brown formula (single and average measures)concept

A formula that predicts how the reliability of a measurement changes when the number of comparable measurements is increased or reduced: ρ_k = k·ρ / (1 + (k - 1)·ρ).

A single item, a single rater or a single measurement carries error. When the average of k parallel measurements is taken, random errors partly cancel out and reliability rises. The Spearman-Brown formula gives the reliability of the average of k measurements when the reliability of one is ρ. It can also be used in reverse to work out how many items or raters are needed to reach a target reliability. In split-half reliability, the correlation between the two halves is stepped up to the full test length with this formula.

It is also the relationship between the 'single measures' and 'average measures' forms of the intraclass correlation coefficient. The formula assumes that the added measurements are of the same quality as the existing ones (parallel); if new items are weaker or measure something different, the prediction is too optimistic. Gains also diminish as reliability rises: approaching very high reliability takes more and more measurements.

An everyday example

In a competition, the scores of a single judge have a reliability of 0.60. Averaging three judges raises it to 3 x 0.60 / (1 + 2 x 0.60) = 1.8 / 2.2 ≈ 0.82. Reaching 0.90 would take six judges.

In MerQur

In the Intraclass Correlation Coefficient analysis, the 'Unit' option: 'Single measures' gives the reliability of one rater and 'Average measures' that of the mean of k raters (Spearman-Brown). In Cronbach's Alpha and Reliability Analysis, 'Additional outputs: Other reliability models: split-half + Guttman λ1-λ6' reports split-half reliability with the Spearman-Brown and Guttman coefficients.

Analyses where this concept appears

Cronbach's Alpha and Reliability AnalysisIntraclass Correlation Coefficient (ICC)

Related concepts

Split-half reliability and Guttman's λIntraclass correlation (ICC)