Skip to content

dark-peak-analytics/darkpeak

Repository files navigation

darkpeak

GitHub release GitHub last commit GitHub top language GitHub repo size R-CMD-Check Coverage

This package was created by Robert Smith & Paul Schneider at the University of Sheffield and Dark Peak Analytics to speed up time dependent markov models.

Installation

Install the development version of the package using the devtools package.

remotes::install_github("RobertASmith/darkpeak")

Quick start

Load the package.

library(darkpeak)
  • ArmaMarkovTDLoop uses C++ to speed up the running of time dependent cohort state transition models (where the transition probability matrix changes over time).
# C++ code

arma::mat ArmaTDMarkovLoop(arma::mat m_TR, arma::cube& a_P )
    {
        int rows = m_TR.n_rows;

        for(int i = 1; i < rows; i++){
            m_TR.row(i) = m_TR.row(i-1) * a_P.slice(i-1);
        }

        return m_TR;
    }
    

Run using the following:

# where:
# m_TR is a markov trace
# a_P is a time dependent transmission array dim = 3.

ArmaTDMarkovLoop(m_TR, a_P)

This results in a significantly faster model run than the equivalent in base R, the below is for a state transition model with 4 health states run over 1000 periods. The units are milliseconds.

Method minimum mean maximum
darkpeak::ArmaTDMarkovLoop 48.2 58.2 175.6
Base R 160.4 197.9 391.2

The Rcpp version runs approximately in approximately 30% of the time of the Base R version, which can make a material difference (e.g. when incorporated into a shiny app a reduction in run-time from 30 seconds to 9 seconds may avoid user frustration, or a model with a 2 hour run-time running in just over 30mins may make it feasible for a meeting).

  • There are plot functions for CEAC plots, CE-Plane plots, plots to check the stability of PSA results, a rankogram for cases where n_strategies >2.
  • There are colour schemes (for all visuals).
  • There is a markdown templates for reporting results (with header and logo etc).
  • Also data for teaching (making markov models shiny parameters?).

Robert Smith 1,2 , Paul Schneider 1,2

1 Public Health Economics and Decision Science, Wellcome Trust Doctoral Training Center, ScHARR, University of Sheffield, UK
2 Dark Peak Analytics, Sheffield, UK

Contact: [email protected]

Website: Dark Peak Analytics

Funding

Rob & Paul are joint funded by the Wellcome Trust Doctoral Training Centre in Public Health Economics and Decision Science [108903] and the University of Sheffield.

About

No description or website provided.

Topics

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages