resvsyhat([varNo,] [usehii:T or F] [graphics keyword phrases]), varNo a positive integer <= ncols(RESIDUALS) |

resvsyhat([graphics keyword phrases]) plots standardized residuals against fitted or predicted values. resvsyhat(usehii:F [,graphics keyword phrases]) does the same without using leverages HII in standardizing. The default is to use HII. resvsyhat(standres:F [,graphics keyword phrases]) does the same without any standardization. The residuals are from variable RESIDUALS or WTDRESIDUALS produced by the most recent GLM (generalized linear or linear model) command such as regress(), anova(), or poisson(). If the most recent command was manova(), only column 1 of the residual matrix is plotted. resvsyhat() cannot be used to make a residual plot after an ARIMA fit using macro arima(). resvsyhat(varNo [, usehii:T or F] [, standres:F] [,graphics keyword phrases]), where varNo is an integer between 1 and ncols(RESIDUALS), plots residuals associated with variable varNo against case numbers. varNo > 1 is legal only when RESIDUALS was computed by manova(). The default plotting symbol is the same as for plot(), a drawn asterisk or star ("\6"). You can change it by including 'symbols:c' as an argument, where c is a CHARACTER or integer scalar or vector. c = 0 is special: it is equivalent to c = "###" and results in points being labeled with case number. See chplot(), subtopic 'symbols_used'. You can use all the usual graphics keywords to modify the default plot characteristics. These include 'title', 'xlab', 'ylab', 'symbols' 'impulse' and 'lines'. See topics 'graphs', 'graph_keys', 'graph_border' and 'graph_ticks'. When you have set option 'dumbplot' to False (see 'options'), the plot will be a low resolution plot unless 'dumb:F' is an argument. Without standres:T, the quantities plotted are r[i]/sd[i] where r[i] is RESIDUALS[i] or WTDRESIDUALS[i] and sd[i] is the estimated standard deviation. WTDRESIDUALS[i] is used after regress(), anova(), or manova() with 'weights:wts' or after nonlinear GLM commands such as logistic() and poisson(). When usehii is True (the default after GLM commands), sd[ii] = sqrt(mse*(1-HII[i])), where mse is the residual mean square after regress(), anova() or manova() or the mean error deviance after non-linear GLM commands. When usehii is False, sd[i] = sqrt(mse). With standres:F, the quantities plotted are r[i]. The values on the X-axis are the estimated means of the response variable. After a nonlinear GLM command, they are in the original scale, not the transformed scale. resvsyhat() is implemented as macro. See also topics resvsindex(), resvsrankits(), resid(), yhat().

Gary Oehlert 2003-01-15