Ridge regression is presented as an analytic technique to be used when predictor variables in a multiple linear regression situation are highly correlated, a situation which may result in unstable regression coefficients and difficulties in interpretation. Ridge regression avoids the problem of selection of variables that may occur in stepwise procedures. Ridge regression is performed by adding a constant to the diagonal of the correlation matrix. This may be done iteratively for a range of constants, say from 0 to 3 in small increments. The results of data analysis are presented both in tables and in graphs, to show how ridge regression reduces the instability of regression weights in cross valdidation. These data were generated to predict fourth and seventh grade achievement from achievement test scores and teacher variables. (Author/CTM)