Check for conditional heteroskedasticity - r

Question

Data:
dput(head(mydata))
structure(list(DATE = structure(c(-315619200, -312940800, -310435200,
-307756800, -305164800, -302486400), tzone = "UTC", class = c("POSIXct",
"POSIXt")), RF = c(0.33, 0.29, 0.35, 0.19, 0.27, 0.24), RMRF = c(-6.99,
0.99, -1.46, -1.7, 3.08, 2.09), SMB = c(2.13, 0.71, -0.65, 0.32,
1.42, -0.24), UMD = c(-3.28, 3.59, 1.85, 2.6, 4.77, 1.03), HML = c(2.65,
-2.15, -2.69, -2.22, -3.83, -0.3), JANDUM = c(1, 0, 0, 0, 0,
0), R4 = c(-4.57, 1.5, -2.83, -1.98, 3.54, 2.15)), row.names = c(NA,
-6L), class = c("tbl_df", "tbl", "data.frame"))
My data contains:
R4 is the percentage return of a portfolio, RF is the return of a good without risk (risk free rate), RMRF is the excess return of portfolio Market Portfolio, SMB, UMD, and HML are 3 factors, and JANDUM is a dummy variable for January (January Dummy).
The data is on a monthly frequency from 1/1960 to 12/2003 (there are 528 observations totaly).
I build the following code for the purpose of "Regress portfolio excess return (R4-RF) to one constant and all other variables (RMRF, SMB, UMD, HML, and JANDUM)".
mydata$PER <-mydata$R4 - mydata$RF
mydata$JANDUM <- as.factor(mydata$JANDUM)
# Fit regression model
model <- lm(PER ~ DATE + RMRF + SMB + UMD + HML + JANDUM, data = mydata)
summary(model)
Then i want to check for the presence of autocorrelation and conditional
heteroskedasticity.
My try:
# // Tests that comes to establish the presence or absence of heteroscedasticity
# Breusch-Pagan test
lmtest::bptest(model)
# Breusch-Pagan test
car::ncvTest(model)
PER_BCMod <- caret::BoxCoxTrans(mydata$PER)
print(PER_BCMod)
# The model for creating the box-cox transformed variable is ready. Lets now apply it on mydata$PER and append it to a new dataframe.
mydata <- cbind(mydata, PER_New = predict(PER_MCMod, mydata$PER))
head(mydata)
# The transformed data for our new regression model is ready. Lets build the model and check for heteroscedasticity.
model_bc <- lm(PER_New ~ DATE + RMRF + SMB + UMD + HML + JANDUM, data=mydata)
bptest(model_bc)
plot(model_bc)
Errors :
> print(PER_BCMod)
Box-Cox Transformation
528 data points used to estimate Lambda
Input data summary:
Min. 1st Qu. Median Mean 3rd Qu. Max.
-30.0700 -2.7800 1.0500 0.6758 4.4325 23.5800
Lambda could not be estimated; no transformation is applied
> # The model for creating the box-cox transformed variable is ready. Lets now apply it on mydata$PER and append it to a new dataframe.
> mydata <- cbind(mydata, PER_New = predict(PER_MCMod, mydata$PER))
Error in predict(PER_MCMod, mydata$PER) : object 'PER_MCMod' not found
I cant understand whats the fault here ? Your suggestions would be welcome.
Is this process wrong , i must follow other solution ?