In christianolms/ContextVariableCreation: Create context variables of a data.frames to enable the analysis of panel data with ordinary least square regression

Dataset

This packages contains the Fatalities dataset. The following code produces an overview of the columns.

library(ContextVariableCreation)
library(data.table)
library(foreach)
library(ggplot2)
data("Fatalities", package = "AER")
#summary(Fatalities)

The dataset contains 7 rows for every of the 48 states for the years 1982 to 1988. First, add traffic death per 1000 miles.

Ordinary least square model

Fatalities$frate <- with(Fatalities, fatal/pop * 10000)

Then create a ordinary least square model.

mod_sig = frate~ beertax
df_reg = create_context_variables_for_panel(Fatalities, group_variable = "state",time_dep_variables = c("beertax"))
m1 = glm(formula = mod_sig, data = df_reg, family = gaussian())
summary(m1)
source("../R/plots.R")
print(plot_residuals(residuals = m1$residuals, target = Fatalities$frate))

Effect of countries

ggplot(aes(x=frate,y=beertax, color=state), data = Fatalities)+geom_point()

Interpretation with beertax in us dollar: if beer tax get higher, the slope

Linear regression with fixed effects

m2 = glm(formula = frate~ beertax + beertax_cv, data = df_reg, family = gaussian())
summary(m2)
print(plot_residuals(residuals = m2$residuals, target = Fatalities$frate))

Model comparision

anova(m1,m2)

christianolms/ContextVariableCreation documentation built on May 13, 2019, 6:57 p.m.