libpulsatile

The development repository for the C++ backend to the unified Bayesian pulsatile hormone modeling algorithm used in the bayespulse R package.

This library extends (and refactors) the single-subject model to multiple subjects and up to two-hormones per patient (driver and response hormones).

The package is currently in development, with only the single subject model currently functioning. Feel free to take a look around and test it out.

---

Dependencies

This library is intended to be used from within an R package, and therefore relies on R's RNGs and datastructures. The list of dependencies for the development phase is rather specific and will be reduced later.

• C++11
• clang (>= 4.0) (on Mac for OpenMP support)
  • libclang-x.x-dev (ubuntu)
  • libomp-dev (Ubuntu)
• R (>= 3.4.3)
• R packages
  • Rcpp
  • RcppArmadillo
  • RInside

Important Note for Cloning this repo

This project structure relies on symbolic links. If you are working on Linux or Mac then you should have not problem cloning and working with this repo. If, however, you are working on a Windows machine there are a few additional steps you will need to take.

1. Windows Vista or newer with NTFS file system, not FAT.
2. You need administrator rights or at least SeCreateSymbolicLinkPrivilege privilege
3. git bash version 2.10.2 or later. It will be helpful to install with core.symlinks option turned on.

In the git bash shell clone the repo. (The example below uses SSH, change the URL as needed for https.)

git clone -c core.symlinks=true git@github.com:BayesPulse/libpulsatile.git

If you are using Windows and you are not sure that you cloned the repo as noted above then please reclone the repo!

Installing R package

From the command line After cloning the repository navigate to the R-package directory and use the provided Makefile. Note that gawk is required on MacOS.

cd <path-to-libpulsatile>/R-package
make install

Within an Interactive R Session

setwd("<path-to-libpulsatile>/R-package")
devtools::load_all(recompile = TRUE)
devtools::document()
devtools::install()

Using RStudio Either use the instructions for installing the package in an interactive R session or open the project via the file bayespulse.Rproj and use the build and install menu options. The project file should tell RStudio to use the Makefile.

Building the C++ binaries

This is only necessary if you are running the binary directly. To build the C++ binary, first ensure the dependencies are satisfied, then build the executable and the test executable with make. To run the tests, run ./bin/tests or ./bin/tests -s to view the detailed test results.

make clean
make
./bin/tests -s

Notes

When compiled outside of the R package, the library is dependent on RInside for random number generators. The chains class also uses R object types and is geared towards structuring the R return object. Gist is, use the R package and C++ only is for dev work.