Aim


ReScience: Repetition, Reproduction & Replication #

The terms repetition, reproduction and replication are not used consistently used. For scientific experiments they are distinguashable by the following properties: team and experimental setup (location, procedure, instrument, conditions).

A Repetition is carried out by the same team using the same setup.

A Reproduction is carried out by different individuals using the same experimental setup. I.e. running the same software on the same input data.

A Replication of a pulished result involves writing and running new code based on the decription provided in the original publication aiming at obtaining the same results.

This followes the definition of Rougier et al. (2017).


Replication principles #

  The replication is reproducible

  The replication is extendable

  The replication tries to implement best practice recommendations

  The replication is auditable


This can be acchieved with the following:

  Reproducibility measures #

  Basic #

  • Use seeds
  • State dependencies and computational environment

  Extended #

  • Run the simulation in a docker container
  • Add a docker file

  Extendability measures #

  Basic #

  • No magic numbers
  • Function based workflow
  • All simulation parameters should be changable

  Extended #

  • Code has provisions for user generated functions (e.g. for different distributions)
  • Package format

  Best practice implementation #

  Basic #

  • Style guide adherence
  • Code commenting

  Extended #

  • Research compendium as R package
  • Unit tests
  • Shiny app for result visualization