Business Glossary

term used in the statistical measurement of relationships within a series. It is one of the assumptions required in a regression in order to make it reliable, also called serial correlation. It means that the error terms are independent of each other [see (a) below]. That is, the deviation of one point about the line (i.e., the error = y- y') is unrelated to the deviation of any other point. When autocorrelation exists [i.e., the error terms are not independent see (b) below], the standard errors of the regression coefficients are seriously underestimated. The problem of autocorrelation is usually detected by the durbin-watson statistic . See illustration.