correlation, split-half

n. The product-moment correlation between paired scores for a group of subjects on each half of a test, which was often used to estimate the internal consistency reliability of the test. This statistic underestimates true internal consistency because of attenuation and includes random error because of differences in the many possible ways of splitting a test in half. Its use has largely been supplanted by use of Kuder-Richardson formula 20 and Cronbach's alpha coefficient.