Can someone explain why the attached R-Studio Code, provided by the author, does not create the output like in his book?

Hi,

I'm afraid I have bad news.

I went through the predictions section of the code and I found everything is as it is supposed to be. The predictions are correct. I tried predictions one step ahead and 59 steps ahead (since the code was already there) and none of them resulted in the plot printed in the book.

Moreover, for these kind of models it is very common for predictions to become "flat" (not entirely constant but almost constant), as in the plot you presented.

Now, when preparing the data.table 'frc.plot'. I had to make some small changes to actually get the final plot.

In other words, as it is, the code is not completely functional to me. These are the small changes I had to make:

Change 1. frc.prod[["ar"]] is a list containing two elements:

frc.prod[["ar"]]$ \$ $pred and frc.prod[["ar"]]$ \$ $se

We don't need the second one. And we must get rid of it when we want to create the data.table 'frc.plot', since we are binding the lists of predictions. I solved this by simply assigning:

frc.prod[["ar"]] <- frc.prod[["ar"]]$pred

Change 2. The predictions are 'ts' objects, but the original Y data is stored as a 'numeric' object. I used as.numeric to change the class of the predictions into numeric vectors.

=================================
The entirety of the plot code I used is here:

# Plot
frc.plot <- data.table('Y' = prod$frc, 'tag' = 'Y')

frc.prod[["ar"]] <- frc.prod[["ar"]]$pred

for (model in c('ar', 'tar', 'star')) {
print(model)
frc.plot <- rbindlist(list(frc.plot,
data.table('Y' = as.numeric(frc.prod[[model]]),
'tag' = paste0('Y_', toupper(model)))))
}
frc.plot[tag == 'Y_AR', tag := 'Y_LINEAR']

ggplot(frc.plot, aes(x = rep(1:59, 4), y = Y, linetype = tag)) +
geom_line() + xlab('') + ylab('') + theme(legend.title = element_blank())

=================================

Now, to answer your question of why those two plots are so different.

One thing I notice is that the plot in the book is made using the base R plot library, whereas the plot in the code provided by the author is made using the ggplot library.

These two different libraries function very differently, and the R object plotted on base R is 100% different to the R object plotted with ggplot. In other words, whatever object the author plotted for the book, it is not what he is plotting in the code he provided.