#Universidade Federal de São Carlos (UFSCar) - Campus Sorocaba
#Departamento de Economia - DEc-So
######################## ALFA CRÍTICO = 5% = 0.05 PARA TODOS OS TESTES ########################
# Preparo dos pacotes
rm(list=ls())
if (!require("pacman")) install.packages("pacman")
## Carregando pacotes exigidos: pacman
## Warning: package 'pacman' was built under R version 4.2.3
library(pacman)
carregar_pacotes <- function() {
pacman::p_load(
readxl,
tseries,
psych,
faraway,
zoo,
lmtest,
stargazer,
texreg,
broom,
knitr,
sandwich,
forecast,
xtable,
forecast
)
}
carregar_pacotes()
# Importação de dados
# Importação da planilha
dados <- read_xlsx("~\\TF_Grupo11\\Dados\\dados.xlsx", sheet = "dados", col_names = TRUE)
# Tratamento de dados
# Definição das variáveis lineares
inpc <- as.vector(dados$inpc)
parroz <- as.vector(dados$parroz)
pfeijao <- as.vector(dados$pfeijao)
pmilho <- as.vector(dados$pmilho)
psoja <- as.vector(dados$psoja)
# Checagem da estacionariedade das variáveis lineares
adf.test(inpc)
##
## Augmented Dickey-Fuller Test
##
## data: inpc
## Dickey-Fuller = -0.96988, Lag order = 4, p-value = 0.9394
## alternative hypothesis: stationary
adf.test(parroz)
##
## Augmented Dickey-Fuller Test
##
## data: parroz
## Dickey-Fuller = -2.5807, Lag order = 4, p-value = 0.3354
## alternative hypothesis: stationary
adf.test(pfeijao)
##
## Augmented Dickey-Fuller Test
##
## data: pfeijao
## Dickey-Fuller = -2.6896, Lag order = 4, p-value = 0.2902
## alternative hypothesis: stationary
adf.test(pmilho)
##
## Augmented Dickey-Fuller Test
##
## data: pmilho
## Dickey-Fuller = -2.0341, Lag order = 4, p-value = 0.5624
## alternative hypothesis: stationary
adf.test(psoja)
##
## Augmented Dickey-Fuller Test
##
## data: psoja
## Dickey-Fuller = -1.6893, Lag order = 4, p-value = 0.7056
## alternative hypothesis: stationary
kpss.test(inpc)
## Warning in kpss.test(inpc): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: inpc
## KPSS Level = 2.3635, Truncation lag parameter = 4, p-value = 0.01
kpss.test(parroz)
## Warning in kpss.test(parroz): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: parroz
## KPSS Level = 1.8001, Truncation lag parameter = 4, p-value = 0.01
kpss.test(pfeijao)
## Warning in kpss.test(pfeijao): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: pfeijao
## KPSS Level = 1.3238, Truncation lag parameter = 4, p-value = 0.01
kpss.test(pmilho)
## Warning in kpss.test(pmilho): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: pmilho
## KPSS Level = 1.7567, Truncation lag parameter = 4, p-value = 0.01
kpss.test(psoja)
## Warning in kpss.test(psoja): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: psoja
## KPSS Level = 1.8107, Truncation lag parameter = 4, p-value = 0.01
# Estudando graficamente
plot(inpc)
plot(parroz)
plot(pfeijao)
plot(pmilho)
plot(psoja)
# Definição das variáveis não-lineares
linpc <- as.vector(log(dados$inpc))
lmilho <- as.vector(log(dados$pmilho))
lsoja <- as.vector(log(dados$psoja))
larroz <- as.vector(log(dados$parroz))
lfeijao <- as.vector(log(dados$pfeijao))
# Checagem da estacionariedade das variáveis não-lineares
adf.test(linpc)
##
## Augmented Dickey-Fuller Test
##
## data: linpc
## Dickey-Fuller = -1.597, Lag order = 4, p-value = 0.7439
## alternative hypothesis: stationary
adf.test(larroz)
##
## Augmented Dickey-Fuller Test
##
## data: larroz
## Dickey-Fuller = -2.6761, Lag order = 4, p-value = 0.2958
## alternative hypothesis: stationary
adf.test(lfeijao)
##
## Augmented Dickey-Fuller Test
##
## data: lfeijao
## Dickey-Fuller = -2.7633, Lag order = 4, p-value = 0.2596
## alternative hypothesis: stationary
adf.test(lmilho)
##
## Augmented Dickey-Fuller Test
##
## data: lmilho
## Dickey-Fuller = -2.582, Lag order = 4, p-value = 0.3348
## alternative hypothesis: stationary
adf.test(lsoja)
##
## Augmented Dickey-Fuller Test
##
## data: lsoja
## Dickey-Fuller = -1.8521, Lag order = 4, p-value = 0.638
## alternative hypothesis: stationary
kpss.test(linpc)
## Warning in kpss.test(linpc): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: linpc
## KPSS Level = 2.3777, Truncation lag parameter = 4, p-value = 0.01
kpss.test(larroz)
## Warning in kpss.test(larroz): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: larroz
## KPSS Level = 1.9522, Truncation lag parameter = 4, p-value = 0.01
kpss.test(lfeijao)
## Warning in kpss.test(lfeijao): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lfeijao
## KPSS Level = 1.3004, Truncation lag parameter = 4, p-value = 0.01
kpss.test(lmilho)
## Warning in kpss.test(lmilho): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lmilho
## KPSS Level = 1.8856, Truncation lag parameter = 4, p-value = 0.01
kpss.test(lsoja)
## Warning in kpss.test(lsoja): p-value smaller than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lsoja
## KPSS Level = 1.8941, Truncation lag parameter = 4, p-value = 0.01
# Estudando graficamente
plot(linpc)
plot(larroz)
plot(lfeijao)
plot(lmilho)
plot(lsoja)
# Regressões:
# LIN-LOG
linlog <- lm(inpc ~ lmilho + lsoja + larroz + lfeijao)
summary(linlog) # Modelo enviesado --> Variáveis não-estacionárias
##
## Call:
## lm(formula = inpc ~ lmilho + lsoja + larroz + lfeijao)
##
## Residuals:
## Min 1Q Median 3Q Max
## -861.46 -235.35 -35.19 296.99 710.17
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -5298.5 769.4 -6.887 3.21e-10 ***
## lmilho 1309.9 193.1 6.784 5.35e-10 ***
## lsoja 219.1 251.3 0.872 0.385104
## larroz 1105.1 303.2 3.645 0.000403 ***
## lfeijao -861.2 178.6 -4.821 4.42e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 337.6 on 115 degrees of freedom
## Multiple R-squared: 0.8349, Adjusted R-squared: 0.8291
## F-statistic: 145.3 on 4 and 115 DF, p-value: < 2.2e-16
# LOG-LOG
loglog <- lm(linpc ~ lmilho + lsoja + larroz + lfeijao)
summary(loglog) # Modelo enviesado --> Variáveis não-estacionárias
##
## Call:
## lm(formula = linpc ~ lmilho + lsoja + larroz + lfeijao)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.20063 -0.05661 -0.00374 0.05595 0.15524
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6.583346 0.165055 39.886 < 2e-16 ***
## lmilho 0.256289 0.041424 6.187 9.69e-09 ***
## lsoja 0.008808 0.053915 0.163 0.871
## larroz 0.266158 0.065041 4.092 7.97e-05 ***
## lfeijao -0.179278 0.038326 -4.678 7.97e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.07242 on 115 degrees of freedom
## Multiple R-squared: 0.8055, Adjusted R-squared: 0.7987
## F-statistic: 119.1 on 4 and 115 DF, p-value: < 2.2e-16
# Teste de normalidade dos resíduos
# LIN-LOG
jarque.bera.test(linlog$residuals)
##
## Jarque Bera Test
##
## data: linlog$residuals
## X-squared = 2.5161, df = 2, p-value = 0.2842
shapiro.test(linlog$residuals)
##
## Shapiro-Wilk normality test
##
## data: linlog$residuals
## W = 0.98398, p-value = 0.1662
# LOG-LOG
jarque.bera.test(loglog$residuals)
##
## Jarque Bera Test
##
## data: loglog$residuals
## X-squared = 2.0386, df = 2, p-value = 0.3608
shapiro.test(loglog$residuals)
##
## Shapiro-Wilk normality test
##
## data: loglog$residuals
## W = 0.98702, p-value = 0.3091
# Multicolinearidade
# LIN-LOG
vi <- cbind(parroz, pfeijao, pmilho, psoja)
cor(as.matrix(vi))
## parroz pfeijao pmilho psoja
## parroz 1.0000000 0.8533685 0.8663793 0.9086823
## pfeijao 0.8533685 1.0000000 0.8382652 0.8506290
## pmilho 0.8663793 0.8382652 1.0000000 0.9214893
## psoja 0.9086823 0.8506290 0.9214893 1.0000000
# LOG-LOG
vi2 <- cbind(larroz, lfeijao, lmilho, lsoja)
cor(as.matrix(vi2))
## larroz lfeijao lmilho lsoja
## larroz 1.0000000 0.8382806 0.9026071 0.9073520
## lfeijao 0.8382806 1.0000000 0.8433941 0.8154014
## lmilho 0.9026071 0.8433941 1.0000000 0.9234753
## lsoja 0.9073520 0.8154014 0.9234753 1.0000000
# Fator de inflação de variância - FIVj e TOLj
# LIN-LOG
fiv <- vif(vi)
FIV <- as.vector(fiv)
TOL <- as.vector(1 / fiv)
fiv_tol <- round(cbind.data.frame(FIV, TOL), 4)
fiv_tol
## FIV TOL
## 1 6.6819 0.1497
## 2 4.3898 0.2278
## 3 7.1949 0.1390
## 4 10.0018 0.1000
# LOG-LOG
fiv2 <- vif(vi2)
FIV2 <- as.vector(fiv2)
TOL2 <- as.vector(1 / fiv2)
fiv_tol2 <- round(cbind.data.frame(FIV2, TOL2), 4)
fiv_tol2
## FIV2 TOL2
## 1 7.3551 0.1360
## 2 3.9005 0.2564
## 3 8.8519 0.1130
## 4 8.4995 0.1177
# Autocorrelação
# Teste Dubin-Watson (DW)
# LIN-LOG
dwtest(linlog) # Modelo enviesado --> Autocorrelação
##
## Durbin-Watson test
##
## data: linlog
## DW = 0.13487, p-value < 2.2e-16
## alternative hypothesis: true autocorrelation is greater than 0
# LOG-LOG
dwtest(loglog) # Modelo enviesado --> Autocorrelação
##
## Durbin-Watson test
##
## data: loglog
## DW = 0.12875, p-value < 2.2e-16
## alternative hypothesis: true autocorrelation is greater than 0
# Teste Breusch-Godfrey (BG)
# LIN-LOG
bgtest(linlog, order = 1, type = c("Chisq")) # Modelo enviesado --> Autocorrelação
##
## Breusch-Godfrey test for serial correlation of order up to 1
##
## data: linlog
## LM test = 99.689, df = 1, p-value < 2.2e-16
bgtest(linlog, order = 3, type = c("Chisq")) # Modelo enviesado --> Autocorrelação
##
## Breusch-Godfrey test for serial correlation of order up to 3
##
## data: linlog
## LM test = 100.27, df = 3, p-value < 2.2e-16
bgtest(linlog, order = 6, type = c("Chisq")) # Modelo enviesado --> Autocorrelação
##
## Breusch-Godfrey test for serial correlation of order up to 6
##
## data: linlog
## LM test = 100.51, df = 6, p-value < 2.2e-16
# LOG-LOG
bgtest(loglog, order = 1, type = c("Chisq")) # Modelo enviesado --> Autocorrelação
##
## Breusch-Godfrey test for serial correlation of order up to 1
##
## data: loglog
## LM test = 99.141, df = 1, p-value < 2.2e-16
bgtest(loglog, order = 3, type = c("Chisq")) # Modelo enviesado --> Autocorrelação
##
## Breusch-Godfrey test for serial correlation of order up to 3
##
## data: loglog
## LM test = 99.694, df = 3, p-value < 2.2e-16
bgtest(loglog, order = 6, type = c("Chisq")) # Modelo enviesado --> Autocorrelação
##
## Breusch-Godfrey test for serial correlation of order up to 6
##
## data: loglog
## LM test = 100.08, df = 6, p-value < 2.2e-16
# Ajustando autocorrelação
# LIN-LOG
# Aplicando diferenciação
inpc_d <- diff(inpc)
larroz_d <- diff(larroz)
lfeijao_d <- diff(lfeijao)
lmilho_d <- diff(lmilho)
lsoja_d <- diff(lsoja)
linlog_2 <- lm(inpc[-1] ~ inpc_d + larroz_d + lfeijao_d + lmilho_d + lsoja_d)
summary(linlog_2) # Ajuste ruim, mas questão da estacionariedade corrigida
##
## Call:
## lm(formula = inpc[-1] ~ inpc_d + larroz_d + lfeijao_d + lmilho_d +
## lsoja_d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1290.69 -692.61 65.35 441.76 1918.11
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4912.179 108.760 45.165 <2e-16 ***
## inpc_d 7.012 3.239 2.165 0.0325 *
## larroz_d 938.926 1504.858 0.624 0.5339
## lfeijao_d -100.836 828.974 -0.122 0.9034
## lmilho_d -234.383 1454.270 -0.161 0.8722
## lsoja_d 785.783 1221.558 0.643 0.5214
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 808.3 on 113 degrees of freedom
## Multiple R-squared: 0.04814, Adjusted R-squared: 0.006024
## F-statistic: 1.143 on 5 and 113 DF, p-value: 0.3419
# Adição de variáveis talvez melhore o modelo
# Rechecagem da estacionariedade das variáveis lineares
adf.test(inpc_d)
## Warning in adf.test(inpc_d): p-value smaller than printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: inpc_d
## Dickey-Fuller = -4.0602, Lag order = 4, p-value = 0.01
## alternative hypothesis: stationary
adf.test(larroz_d)
## Warning in adf.test(larroz_d): p-value smaller than printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: larroz_d
## Dickey-Fuller = -4.4259, Lag order = 4, p-value = 0.01
## alternative hypothesis: stationary
adf.test(lfeijao_d)
## Warning in adf.test(lfeijao_d): p-value smaller than printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: lfeijao_d
## Dickey-Fuller = -4.3661, Lag order = 4, p-value = 0.01
## alternative hypothesis: stationary
adf.test(lmilho_d)
##
## Augmented Dickey-Fuller Test
##
## data: lmilho_d
## Dickey-Fuller = -3.6597, Lag order = 4, p-value = 0.0309
## alternative hypothesis: stationary
adf.test(lsoja_d)
## Warning in adf.test(lsoja_d): p-value smaller than printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: lsoja_d
## Dickey-Fuller = -4.6323, Lag order = 4, p-value = 0.01
## alternative hypothesis: stationary
kpss.test(inpc_d)
## Warning in kpss.test(inpc_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: inpc_d
## KPSS Level = 0.26578, Truncation lag parameter = 4, p-value = 0.1
kpss.test(larroz_d)
## Warning in kpss.test(larroz_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: larroz_d
## KPSS Level = 0.098779, Truncation lag parameter = 4, p-value = 0.1
kpss.test(lfeijao_d)
## Warning in kpss.test(lfeijao_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lfeijao_d
## KPSS Level = 0.056224, Truncation lag parameter = 4, p-value = 0.1
kpss.test(lmilho_d)
## Warning in kpss.test(lmilho_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lmilho_d
## KPSS Level = 0.12889, Truncation lag parameter = 4, p-value = 0.1
kpss.test(lsoja_d)
## Warning in kpss.test(lsoja_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lsoja_d
## KPSS Level = 0.12128, Truncation lag parameter = 4, p-value = 0.1
# Variáveis estacionárias --> Teórica correção da fonte de autocorrelação
# LOG-LOG
# Aplicando diferenciação
linpc_d <- diff(linpc)
larroz_d <- diff(larroz)
lfeijao_d <- diff(lfeijao)
lmilho_d <- diff(lmilho)
lsoja_d <- diff(lsoja)
loglog_2 <- lm(linpc[-1] ~ linpc_d + larroz_d + lfeijao_d + lmilho_d + lsoja_d)
summary(loglog_2) # Ajuste ruim, mas questão da estacionariedade corrigida
##
## Call:
## lm(formula = linpc[-1] ~ linpc_d + larroz_d + lfeijao_d + lmilho_d +
## lsoja_d)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.277357 -0.120192 0.006906 0.075785 0.299998
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.522045 0.022958 371.194 <2e-16 ***
## linpc_d -0.341902 3.616094 -0.095 0.925
## larroz_d 0.164507 0.302586 0.544 0.588
## lfeijao_d 0.004115 0.166745 0.025 0.980
## lmilho_d 0.169519 0.293802 0.577 0.565
## lsoja_d 0.086052 0.246143 0.350 0.727
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.1624 on 113 degrees of freedom
## Multiple R-squared: 0.01108, Adjusted R-squared: -0.03268
## F-statistic: 0.2531 on 5 and 113 DF, p-value: 0.9375
# Exclusão/adição de variáveis talvez melhore o problema --> R ajustado indica penalização por variável não-significativa
# Rechecagem da estacionariedade das variáveis lineares
adf.test(linpc_d)
## Warning in adf.test(linpc_d): p-value smaller than printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: linpc_d
## Dickey-Fuller = -4.0592, Lag order = 4, p-value = 0.01
## alternative hypothesis: stationary
adf.test(larroz_d)
## Warning in adf.test(larroz_d): p-value smaller than printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: larroz_d
## Dickey-Fuller = -4.4259, Lag order = 4, p-value = 0.01
## alternative hypothesis: stationary
adf.test(lfeijao_d)
## Warning in adf.test(lfeijao_d): p-value smaller than printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: lfeijao_d
## Dickey-Fuller = -4.3661, Lag order = 4, p-value = 0.01
## alternative hypothesis: stationary
adf.test(lmilho_d)
##
## Augmented Dickey-Fuller Test
##
## data: lmilho_d
## Dickey-Fuller = -3.6597, Lag order = 4, p-value = 0.0309
## alternative hypothesis: stationary
adf.test(lsoja_d)
## Warning in adf.test(lsoja_d): p-value smaller than printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: lsoja_d
## Dickey-Fuller = -4.6323, Lag order = 4, p-value = 0.01
## alternative hypothesis: stationary
kpss.test(linpc_d)
## Warning in kpss.test(linpc_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: linpc_d
## KPSS Level = 0.15415, Truncation lag parameter = 4, p-value = 0.1
kpss.test(larroz_d)
## Warning in kpss.test(larroz_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: larroz_d
## KPSS Level = 0.098779, Truncation lag parameter = 4, p-value = 0.1
kpss.test(lfeijao_d)
## Warning in kpss.test(lfeijao_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lfeijao_d
## KPSS Level = 0.056224, Truncation lag parameter = 4, p-value = 0.1
kpss.test(lmilho_d)
## Warning in kpss.test(lmilho_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lmilho_d
## KPSS Level = 0.12889, Truncation lag parameter = 4, p-value = 0.1
kpss.test(lsoja_d)
## Warning in kpss.test(lsoja_d): p-value greater than printed p-value
##
## KPSS Test for Level Stationarity
##
## data: lsoja_d
## KPSS Level = 0.12128, Truncation lag parameter = 4, p-value = 0.1
# Variáveis estacionárias --> Teórica correção da fonte de autocorrelação
# Repetindo o teste Dubin-Watson (DW)
# LIN-LOG
dwtest(linlog_2) # Autocorrelação persistente --> Não tem origem única na estacionariedade
##
## Durbin-Watson test
##
## data: linlog_2
## DW = 0.039475, p-value < 2.2e-16
## alternative hypothesis: true autocorrelation is greater than 0
# LOG-LOG
dwtest(loglog_2) # Autocorrelação persistente --> Não tem origem única na estacionariedade
##
## Durbin-Watson test
##
## data: loglog_2
## DW = 0.013671, p-value < 2.2e-16
## alternative hypothesis: true autocorrelation is greater than 0
# Repetindo o Teste Breusch-Godfrey (BG)
# LIN-LOG
bgtest(linlog_2, order = 1, type = c("Chisq")) # Autocorrelação persistente --> Não tem origem única na estacionariedade
##
## Breusch-Godfrey test for serial correlation of order up to 1
##
## data: linlog_2
## LM test = 113.06, df = 1, p-value < 2.2e-16
bgtest(linlog_2, order = 3, type = c("Chisq")) # Autocorrelação persistente --> Não tem origem única na estacionariedade
##
## Breusch-Godfrey test for serial correlation of order up to 3
##
## data: linlog_2
## LM test = 113.19, df = 3, p-value < 2.2e-16
bgtest(linlog_2, order = 6, type = c("Chisq")) # Autocorrelação persistente --> Não tem origem única na estacionariedade
##
## Breusch-Godfrey test for serial correlation of order up to 6
##
## data: linlog_2
## LM test = 113.29, df = 6, p-value < 2.2e-16
# LOG-LOG
bgtest(loglog_2, order = 1, type = c("Chisq")) # Autocorrelação persistente --> Não tem origem única na estacionariedade
##
## Breusch-Godfrey test for serial correlation of order up to 1
##
## data: loglog_2
## LM test = 115.18, df = 1, p-value < 2.2e-16
bgtest(loglog_2, order = 3, type = c("Chisq")) # Autocorrelação persistente --> Não tem origem única na estacionariedade
##
## Breusch-Godfrey test for serial correlation of order up to 3
##
## data: loglog_2
## LM test = 115.2, df = 3, p-value < 2.2e-16
bgtest(loglog_2, order = 6, type = c("Chisq")) # Autocorrelação persistente --> Não tem origem única na estacionariedade
##
## Breusch-Godfrey test for serial correlation of order up to 6
##
## data: loglog_2
## LM test = 115.25, df = 6, p-value < 2.2e-16
# Heterocedasticidade
# Teste Breusch-Pagan (BP)
# Modelos LIN-LOG
# LIN-LOG
bptest(linlog)
##
## studentized Breusch-Pagan test
##
## data: linlog
## BP = 13.297, df = 4, p-value = 0.009913
# LIN-LOG AJUSTADO
bptest(linlog_2)
##
## studentized Breusch-Pagan test
##
## data: linlog_2
## BP = 4.056, df = 5, p-value = 0.5414
# Correção da heterocedastidade por meio do ajuste
# Modelos LOG-LOG
# LOG-LOG
bptest(loglog)
##
## studentized Breusch-Pagan test
##
## data: loglog
## BP = 14.345, df = 4, p-value = 0.006272
# LOG-LOG AJUSTADO
bptest(loglog_2)
##
## studentized Breusch-Pagan test
##
## data: loglog_2
## BP = 10.065, df = 5, p-value = 0.07341
# Correção da heterocedastidade por meio do ajuste
# Teste de White
# Modelos LIN-LOG
# LIN-LOG
res1 <- (linlog$residuals^2)
arroz_2_n <- larroz^2
feijao_2_n <- lfeijao^2
soja_2_n <- lsoja^2
milho_2_n <- lmilho^2
resq_linlog <- lm(res1 ~ larroz + lfeijao + lsoja + lmilho +
arroz_2_n + feijao_2_n + soja_2_n + milho_2_n +
larroz * lfeijao + larroz * lsoja + larroz * lmilho +
lfeijao * lsoja + lfeijao * lmilho +
lsoja * lmilho
)
greslinlog <- glance(resq_linlog)
Rsq <- greslinlog$r.squared
N <- nobs(resq_linlog)
S <- greslinlog$df
nR2 <- N * Rsq
chisq_vc <- qchisq(0.95, df = (S - 1))
nR2
## [1] 35.4055
chisq_vc
## [1] 22.36203
# LIN-LOG AJUSTADO
res1_2 <- (linlog_2$residuals^2)
arroz_2_d <- larroz_d^2
feijao_2_d <- lfeijao_d^2
soja_2_d <- lsoja_d^2
milho_2_d <- lmilho_d^2
resq_linlog_2 <- lm(res1_2 ~ larroz_d + lfeijao_d + lsoja_d + lmilho_d +
arroz_2_d + feijao_2_d + soja_2_d + milho_2_d +
larroz_d * lfeijao_d + larroz_d * lsoja_d + larroz_d * lmilho_d +
lfeijao_d * lsoja_d + lfeijao_d * lmilho_d +
lsoja_d * lmilho_d
)
greslinlog_2 <- glance(resq_linlog_2)
Rsq_2 <- greslinlog_2$r.squared
N2 <- nobs(resq_linlog_2)
S2 <- greslinlog_2$df
nR2_2 <- N2 * Rsq_2
chisq_vc_2 <- qchisq(0.95, df = (S - 1))
nR2_2
## [1] 13.97686
chisq_vc_2
## [1] 22.36203
# Modelos LOG-LOG
# LOG-LOG
res3 <- (loglog$residuals^2)
larroz_2 <- larroz^2
lfeijao_2 <- lfeijao^2
lsoja_2 <- lsoja^2
lmilho_2 <- lmilho^2
resq_loglog_3 <- lm(res3 ~ larroz + lfeijao + lsoja + lmilho +
larroz_2 + lfeijao_2 + lsoja_2 + lmilho_2 +
larroz * lfeijao + larroz * lsoja + larroz * lmilho +
lfeijao * lsoja + lfeijao * lmilho +
lsoja * lmilho
)
gresloglog_3 <- glance(resq_loglog_3)
Rsq2_3 <- gresloglog_3$r.squared
N3 <- nobs(resq_loglog_3)
S3 <- gresloglog_3$df
nR2_3 <- N3 * Rsq2_3
chisq_vc_3 <- qchisq(0.95, df = (S - 1))
nR2_3
## [1] 35.21795
chisq_vc_3
## [1] 22.36203
# LOG-LOG AJUSTADO
res4 <- (loglog_2$residuals^2)
larroz_2_d <- larroz_d^2
lfeijao_2_d <- lfeijao_d^2
lsoja_2_d <- lsoja_d^2
lmilho_2_d <- lmilho_d^2
resq_loglog_4 <- lm(res4 ~ larroz_d + lfeijao_d + lsoja_d + lmilho_d +
larroz_2_d + lfeijao_2_d + lsoja_2_d + lmilho_2_d +
larroz_d * lfeijao_d + larroz_d * lsoja_d + larroz_d * lmilho_d +
lfeijao_d * lsoja_d + lfeijao_d * lmilho_d +
lsoja_d * lmilho_d
)
gresloglog_4 <- glance(resq_loglog_4)
Rsq2_4 <- gresloglog_4$r.squared
N4 <- nobs(resq_loglog_4)
S4 <- gresloglog_4$df
nR2_4 <- N4 * Rsq2_4
chisq_vc_4 <- qchisq(0.95, df = (S - 1))
nR2_4
## [1] 20.38747
chisq_vc_4
## [1] 22.36203
warnings() # Avisos limitam-se a p-valores menores que o informado