Which versions of R an Python?

[alex-m-ffm]

Hi Daniel,

I am trying to replicate your results, but I am having trouble getting the Python packages gluonts[mxnet] and pyflux to work, as well as the R packages mfbvar, nowcastDFM and nowcastLSTM. The metadata of your Jupyter notebooks indicates Python version 3.9.7 and R 4.1.2, so I tried setting up a conda environment with those but still no luck. gluonts did get installed but then throws an error in the deepvar notebook. Including pyflux in my .yml-file simply breaks down the whole setup so for now I removed it.
Regarding the R packages, I guess I need to install them from Github rather than from CRAN? I can run the install commands but get errors that they are not available for my version of R and google search showed that at least mfbvar and nowcastDFM have been removed from CRAN.
Some information on your setup would be much appreciated. Even better, if you have it, an environment.yml or PIP requirements.txt file.

Many thanks and best regards,
Alex

[dhopp1]

What is your purpose in trying to run them all? DeepVAR (using gluonts) and MF-VAR (Pyflux) didn't perform well, so I wouldn't even bother spending time trying to get those to work. Pyflux especially only works with Python < 3.0 so would need its own environment.

There is no point in using nowcastLSTM as well if you can use Python, it is just an R wrapper for nowcast_lstm meant for R users who don't know Python.

Yes, nowcastDFM and mfbvar need to be installed via Github.

Running from within R:

install.packages("devtools")
devtools::install_github("dhopp1/nowcastDFM")
devtools::install_github("ankargren/mfbvar")

[SermetPekin]

Hi @alex-m-ffm and @dhopp1,

I cloned the repository to reproduce the results for as many notebooks as possible.
I also created some statistics how long they each require to finish, (one can also rerun them easily on their environments with automatic python files that run all of them with some configuration options. Documented all options in the clone)
cleaned a few typos in the code.
here is my clone to reproduce them. Clone

for example these for missing packages

Documentation to reproduce my humble version of the reproduction which includes requirements.txt file for python also

# Install required R packages (automated script - recommended)
Rscript install_r_packages.R

# OR install manually:
# First install CRAN packages
R -e 'install.packages(c("IRkernel", "tidyverse", "plyr", "dplyr", "midasr", "midasml", "imputeTS", "Rmisc", "mfbvar", "devtools"), repos="https://cran.r-project.org")'

# Then install GitHub packages
R -e 'devtools::install_github(c("dhopp1/nowcastDFM", "dhopp1/nowcastLSTM"))'

# Install the R kernel for Jupyter
R -e 'IRkernel::installspec(user = TRUE)'

 @dhopp1 Thanks again for this cool repository that includes all necessary files to reproduce.