regcoefs(Model [,pvals:T] [,byvar:F]) or regcoefs([pvals:T] [,byvar:F]), where Model is a CHARACTER scalar |

regcoefs(Model) returns a matrix with appropriately labeled rows and columns of the regression coefficients, their standard errors and t-statistics from a least squares fit to the regression model specified by Model. There can be no factors in Model. If Model is omitted, the most recent GLM model is used. regcoefs(Model,pvals:T) or regcoefs(pvals:T) also computes two-tail P values corresponding to the t-statistics on the basis of Student's t-distribution with degrees of freedom from the last element of side effect variable DF. Because of the presence of row and column labels, after any GLM command with a model without factors, typing regcoefs([pvals:T]) produces a table similar to that produced by regress(). After non-linear fits such as logistic() or poisson(), the P-values will not necessarily be appropriate. If the response variable is multivariate, the result is a structure, each of whose components is a labeled matrix of coefficients, standard errors and t-statistics. regcoefs(Model,byvar:F) or regcoefs(byvar:F) returns a single labeled matrix, with separate columns for the coefficients, standard errors, ... for each variable. See also topics 'glm', regress(), secoefs().

Gary Oehlert 2003-01-15