Movatterモバイル変換

lgpr

R-package for interpretable nonparametric modeling of longitudinaldata using additive Gaussian processes. Contains functionality forinferring covariate effects and assessing covariate relevances. Variousmodels can be specified using a convenient formula syntax.

[!NOTE] Using this package is computationally viable if your data sethas maybe less than 300 observations. But the much more scalablelgpr2 package has beenreleased! It is much faster but unfortunately doesn’t have all thespecial modeling features included in this package.

Getting started

See overview, tutorials, vignettes and documentation athttps://jtimonen.github.io/lgpr-usage/index.html.

Requirements

• The package should work on all major operating systems.
• R 3.4 or later is required, R 4.2 or later is recommended

Installing from CRAN

• The latest released version that is available from CRAN can beinstalled simply via

install.packages("lgpr")

Installing from CRAN is probably the easiest option since they mighthave binaries for your system (so no need to build the package fromsource yourself).

Installing from source

• The latest released version (which might not be in CRAN yet) can beinstalled via

install.packages('devtools')# if you don't have devtools alreadydevtools::install_github('jtimonen/lgpr',build_vignettes =TRUE)

• The latest development version can be installed via

devtools::install_github('jtimonen/lgpr',ref ="develop")

Github installations are source installations (they require a C++compiler).

• If you have trouble installing the dependencyrstan, seetheseinstructions
• Installing from source requires that you have your toolchain setupproperly. See the instructions for:
  • Linux
  • Windows
  • Mac

Using R < 4.2

If you are usingR version 4.1 or earlier, you can getan error

cc1plus.exe: out of memory allocating 65536 bytesmake: *** [C:/PROGRA~1/R/R-40~1.2/etc/i386/Makeconf:227: stanExports_lgp_latent.o] Error 1

because both 64-bit and 32-bit versions of the package are gettinginstalled. To disable this and resolve error, ugrade to latest R orinstall the version that hasBiarch: false by

devtools::install_github('jtimonen/lgpr',ref ="no-biarch")

Real data andreproducing the experiments

For code to reproduce the experiments of our manuscript seehttps://github.com/jtimonen/lgpr-usage. Preprocessed longitudinalproteomics data is also provided there. See also the built-inread_proteomics_data() function.