regs(x,y [,T] [GLM keywords]), x and y REAL matrices with the same number of rows; T means no intercept |

regs(x,y) computes the regression of y on the columns of x. For example, when data is a n by 7 matrix, say, you can compute the regression of the last column on the first 6 by regs(data[,-7], data[,7]). regs(x,y,T) does the same, except the model fit has no constant term (intercept). You can use GLM keyword phrases such as 'pvals:T', 'silent:T', 'wts:weights', and 'marginal:T' as additional arguments. Macro regs() creates temporary variables @X1, @X2, ... and @Y and then invokes regress() or, if y has more than 1 column, manova(). When y is univariate, immediately follow regs() by anova() to see the ANOVA table. When y is multivariate, follow regs() by regcoefs() to see coefficients and standard errors. Because the model for regress() or manova() uses temporary variables, STRMODEL cannot be used as a model for a subsequent regress(), anova(), or manova() command. However, these variables can be retrieved, by modelvars(). For example, when x has 3 columns, Cmd> makecols(modelvars(x:T),X1,X2,X2) creates variables X1, X2 and X3. Of course, if x still exists, makecols(x,X1,X2,X3) does the same. See also regress(), anova(), modelvars(), regcoefs(), and 'glm_keys'.

Gary Oehlert 2003-01-15