reliability, split-half

n. The level of a test's measurement error determined by examining the consistency of the scores of two halves of a given test. Splithalf reliability is concerned with internal consistency. To calculate split-half reliability, the test is divided in half (division can be done by randomly selecting half the questions, dividing by the first and last half of the items, or by using the odd-even system of item selection), and then each half is scored separately. Unlike other measures of reliability that simply correlate the resultant scores, a simple correlation of the two halves would provide an underestimation of the overall test reliability (as reliability increases with more items). As such, the Spearman-Brown formula is applied to correct for the half-length; the SpearmanBrown formula adjusts the correlation calculations to estimate what the reliability would be if each half of the test had been the length of the entire test.