EVPPI in Windows

Health economics
Software
R
Bayesian statistics
Author

Written by Gianluca

Published

April 9, 2021

A couple of colleagues have alerted us to some issues with the BCEA function evppi, under MS Windows. Basically, it seemed as if under MS Windows and with R 4.0.0 (which is a recent, though not the most recent release of R) the evppi function would break, when trying to make the analysis based on INLA (and the work discussed here).

In particular, if the user didn’t have the R package INLA already installed, when trying to run

x=evppi(..., method="INLA")

BCEA would oblige and “suggest” the installation of INLA. Now this is also a bit complicated, because INLA is a “heavy” package and, in addition, is not installable from the main CRAN. Under Linux and (I think!) Mac OS, this isn’t a problem and BCEA would simply tell the user to install INLA — though the link to the URL from which this can be done was outdated in the current CRAN release of BCEA. Under MS Windows, however, I think that things become a bit more complicated as, in order to install a package not on CRAN, another package is needed (Rtools). But Rtools in its most recent version (which would be aligned with R) wouldn’t install directly. This then would make it impossible to install INLA and its main dependencies and therefore would break evppi.

I did a bit of digging and I found a solution (which my colleagues have confirmed has worked for them!). So, if:
- you are under MS Windows and have a version of R that is at least as recent as 4.0.0;
- you have never installed INLA;
- you have installed BCEA and now want to use evppi to compute the EVPPI for your model, using the fast algorithm based on INLA
here’s what you need to do:

  1. Follow the instructions here to install the most recent version of Rtools;
  2. Install INLA using the following code
install.packages("INLA",repos=c(getOption("repos"),INLA="https://inla.r-inla-download.org/R/stable"), dep=TRUE)
  1. If you don’t already have it installed, install the package ldr (which is needed to complement the fast EVPPI algorithm), using the code
install.packages('ldr')

and then you’re good to go!