The null hypothesis is that the response variable Y is related to k explanatory variables by the multiple regression model E(Y)=Xβ, for which the residual sum of squares (see ANOVA) is R. The data consists of two sub-samples (1 and 2; of sizes n1 and n2) and the alternative hypothesis is that in sub-sample j (j=1,2) the model is E(Y)=Xβj with residual sum of squares Rj. The Chow test, introduced in 1960, assumes independent normal errors and has test statistic F, given by The test statistic is compared with an F-distribution having k and (n1+n2−2k) degrees of freedom.