Which method accounts for item variance in split-half reliability by using the standard deviations of each half?

Which method accounts for item variance in split-half reliability by using the standard deviations of each half?

• A. Spearman-Brown prophecy formula.
• B. Flanagan and Rulon method.
• C. Cronbach's alpha.
• D. Kuder-Richardson 20.

Answer: (B)

Split-half reliability hinges on turning the observed correlation between two halves into an estimate of how reliable the whole test is. When the two halves don’t share the same amount of item variance, you need a correction that reflects that difference. The Flanagan and Rulon method does exactly that by using the standard deviations of each half in the adjustment. By weighting each half by its own variability, it accounts for the unequal item variance and provides a more accurate reliability estimate for the full test. Other approaches rely on different assumptions: the Spearman-Brown correction uses the half-correlation and assumes equal variances; Cronbach’s alpha combines covariances across all items under tau-equivalence; Kuder-Richardson 20 is for binary items and relates to alpha under specific conditions.

Split-half reliability hinges on turning the observed correlation between two halves into an estimate of how reliable the whole test is. When the two halves don’t share the same amount of item variance, you need a correction that reflects that difference. The Flanagan and Rulon method does exactly that by using the standard deviations of each half in the adjustment. By weighting each half by its own variability, it accounts for the unequal item variance and provides a more accurate reliability estimate for the full test.

Other approaches rely on different assumptions: the Spearman-Brown correction uses the half-correlation and assumes equal variances; Cronbach’s alpha combines covariances across all items under tau-equivalence; Kuder-Richardson 20 is for binary items and relates to alpha under specific conditions.