Considere esses exemplos lattice: http://leg.ufpr.br/~walmes/cursoR/mgest/1medidas-influen.html ggplot2: http://leg.ufpr.br/~walmes/analises/CECarducci/cafe_pedotrans.html Usando a graphics com *código reproduzível* m0 <- lm(dist ~ poly(speed, degree = 2), data = cars) grid <- with(cars, data.frame(speed = seq(min(speed), max(speed), length.out = 31))) mat <- predict(m0, newdata = grid, interval = "confidence") grid <- cbind(grid, as.data.frame(mat)) str(grid) plot(dist ~ speed, data = cars, type = "n") with(grid, polygon(x = c(speed, rev(speed)), y = c(lwr, rev(upr)), col = "gray", border = NA)) lines(fit ~ speed, data = grid) points(dist ~ speed, data = cars) À disposição. Walmes.