R - Advanced Regression Models

Ad: Free Python Tutorials | Free R Tutorials | Free Deep Learning Tutorials | Free Machine Learning Tutorials | Free Artificial Intelligence Tutorials

Register now on www.gladwinanalytics.com to get started with 7500+ hours of free video tutorials on data science

----------------------------------------------------------------------------------

Each of the regression analysis below contains working code examples with brief use-case explanations covered for each of the regression types in the list below. Many of these code snippets are generic enough so you could use them as a base template to start and build up on for your analyses.

Please note that the information presented in here should not be construed as full and complete analysis, but rather as a template and a hand guide of available modeling options. You are advised to pursue independent and thorough research before arriving at conclusions.

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 freedom

Compare 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 model

Calculate 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 0.1317191

As expected, the errors from the robust regression model is lesser than the linear regression model.

Learn more

Logistic Regression

Probit Regression

Multinomial Regression

Ordinal Logistic

Poisson and Negative Binomial

Ridge Regression

Beta Regression

Dirichlet Regression

Loess Regression

Isotonic Regression

---------------------------------------------------------------------------------

Ad: Free Python Tutorials | Free R Tutorials | Free Deep Learning Tutorials | Free Machine Learning Tutorials | Free Artificial Intelligence Tutorials

Register now on www.gladwinanalytics.com to get started with 7500+ hours of free video tutorials on data science

----------------------------------------------------------------------------------