Blog5: Dummy Variable Regression

## load library
if (!require("fastDummies")) install.packages("fastDummies")
if (!require("tidyverse")) install.packages("tidyverse")# Metapackge

Introduction

Dummy variables (or binary variables) are commonly used in statistical analyses and in more simple descriptive statistics. A dummy column is one which has a value of one when a categorical event occurs and a zero when it doesn’t occur. For example, if we had a dummy variable called male, 1 would indicate that the individual is male and 0 would indicate that the individual is female (or non-male to be more precise).

Import Data

set.seed(1)
# Import training data
train <- read.csv('https://raw.githubusercontent.com/Vinayak234/DATA605/master/train.csv')
train <- train %>%
  select(MSZoning, LotArea, SalePrice)
head(train)

Create dummy variables

results <- fastDummies::dummy_cols(train,remove_first_dummy = TRUE)
head(results)

Dummy variables as independent variables

If the data come as predefined dummy variables, then it is rather straightforward to use these in regressions.

m1<- lm(SalePrice~.-MSZoning,data=results)
summary(m1)

## 
## Call:
## lm(formula = SalePrice ~ . - MSZoning, data = results)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -235390  -45628  -12485   26231  546762 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 5.756e+04  2.320e+04   2.481   0.0132 *  
## LotArea     1.786e+00  1.960e-01   9.110  < 2e-16 ***
## MSZoning_FV 1.446e+05  2.484e+04   5.820 7.22e-09 ***
## MSZoning_RH 6.082e+04  2.948e+04   2.063   0.0393 *  
## MSZoning_RL 1.128e+05  2.323e+04   4.856 1.33e-06 ***
## MSZoning_RM 5.736e+04  2.365e+04   2.425   0.0154 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 73120 on 1454 degrees of freedom
## Multiple R-squared:  0.1558, Adjusted R-squared:  0.1528 
## F-statistic: 53.65 on 5 and 1454 DF,  p-value: < 2.2e-16

Conclusion

As we just showed, linear regression is not strictly limited to quantitative variables. We can easily use least squares to create a regression model with qualitative explanatory variables. Futhermore, while it isn’t required that dummy variables take on the values of 1 and 0, it makes interpretation of your models much easier. We can extend linear regression with dummy variables a bit further, but let’s save that for another time.