Robust Regression
Robust regression can be used in any situation where OLS regression can be applied. It generally gives better accuracies over OLS because it uses a weighting mechanism to weigh down the influential observations. It is particularly resourceful when there are no compelling reasons to exclude outliers in your data.
Robust regression can be implemented using the rlm() function in MASS package. The outliers can be weighted down differently based on psi.huber, psi.hampel and psi.bisquare methods specified by the psi argument.
How To Specify A Robust Regression Model
library(MASS)
rlm_mod <- rlm(stack.loss ~ ., stackloss, psi = psi.bisquare) # robust reg model
summary(rlm_mod)
#> Call: rlm(formula = stack.loss ~ ., data = stackloss)
#> Residuals:
#> Min 1Q Median 3Q Max
#> -8.91753 -1.73127 0.06187 1.54306 6.50163
#>
#> Coefficients:
#> Value Std. Error t value
#> (Intercept) -41.0265 9.8073 -4.1832
#> Air.Flow 0.8294 0.1112 7.4597
#> Water.Temp 0.9261 0.3034 3.0524
#> Acid.Conc. -0.1278 0.1289 -0.9922
#>
#> Residual standard error: 2.441 on 17 degrees of freedomCompare Performance of rlm() with lm()
Lets build the equivalent lm() model so we can compare the errors against the respective fitted values.
lm_mod <- lm(stack.loss ~ ., stackloss) # lm reg modelCalculate the Errors
# Errors from lm() model
DMwR::regr.eval(stackloss$stack.loss, lm_mod$fitted.values)
#> mae mse rmse mape
#> 2.3666202 8.5157125 2.9181694 0.1458878
# Errors from rlm() model
DMwR::regr.eval(stackloss$stack.loss, rlm_mod$fitted.values)
#> mae mse rmse mape
#> 2.1952232 9.0735283 3.0122298 0.1317191As expected, the errors from the robust regression model is lesser than the linear regression model.
