#Set working directory
setwd("C:/Andrzej/Dydaktyka/econometric_methods/")
#Import dataset from CSV
dataset <- read.csv("lecture_5_6_sc.csv", header = TRUE, sep = ",", dec = ".")

#PART 1
OLS_results <- lm(log_Q_gasoline ~ log_P_gasoline + log_P_new_car + log_P_used_car + log_Income, data = dataset)
summary(OLS_results)

#PART 2
install.packages("lmtest")
library(lmtest)
dwtest(OLS_results,alternative="two.sided")  # "positive", "negative"

#Exercise: write a script to perform the h-Durbin test in the model additionally containing the lagged dependent variable.

bgtest(OLS_results, order = 4)

install.packages("stats")
library(stats)
Box.test(OLS_results$residuals, lag = 4, type = "Ljung-Box")
acf(OLS_results$residuals, lag.max = 12)
pacf(OLS_results$residuals, lag.max = 12)


#PART 3
install.packages("sandwich")
library(sandwich)
Newey.West.Matrix <- vcovHAC(OLS_results)
coeftest(OLS_results, df = OLS_results$df.residual, vcov = Newey.West.Matrix)

install.packages("orcutt")
library(orcutt)
cochrane.orcutt(OLS_results)

#Exercise 1: test for serial correlation in the estimated production function (data: lecture_5_6_ex1.csv), 
#in which the log value added (l_valueadd) depends on log labour (l_labour) and log capital (l_capital).
#The model includes a constant. If serial correlation is present, attempt to overcome the problem.

#Exercise 2: test for serial correlation in the estimated Phillips curve model with static inflation expectations
# (data: lecture_5_6_ex2.csv), in which the inflation rate (d_infl) depends on the unemployment rate (unempl).
#The model includes a constant. If serial correlation is present, attempt to overcome the problem.