REPROdUCIBLE {{ ANALYSIS }} IN R
Safa Andac
GENERATING REPORTS IN R
Safa Andac
x100??
Expectation
x1
Reality
x10
openclipart.com
WRITING CODE AT FIRST
Import
Tidy
Transform
Model
Visualize
Communicate
Exploration
WORKFLOW
Import
Tidy
Transform
Model
Visualize
Communicate
Exploration
WORKFLOW
Import
Tidy
Transform
Model
Visualize
Communicate
Exploration
WORKFLOW
DILEMMA
Reproduce or Save Time?
openclipart.com
ALL FOR ONE
divide.R
and.R
conquer.R
the.R
world.R
- Time Saver
- Easier to Debug
- Reproducibility comes at a cost
- Responsible for the code base
ONE FOR ONE
conquer_the_world.R
- Time Consuming
- Hard to Debug
- Reproducible (generally)
- Responsible for one file
~x500
SOLUTION
Use a pipeline tool
- Time Saver
- Keeps tracks of changes
- Supports reproducilibity
SOLUTION
Imperative Scripts
Embrace Functions!
SOLUTION
Workflow as a list of targets
tar_visnetwork()
SOLUTION
tar_visnetwork()
tar_make()
SOLUTION
Build the pipeline gradually:
- Add target
- Run Pipeline
- Inspect New Targets
- Repeat
ONE MORE THING
Option 1 (?)
Option 2
Option 3
EMBRACE THE EMBRACE!
References
- https://github.com/ropensci/targets
- https://wlandau.github.io/targets-tutorial
- https://github.com/tidyverse/tidyverse
- https://rlang.r-lib.org/reference/topic-data-mask.html
- https://rlang.r-lib.org/reference/topic-data-mask-programming.html
Reproducible Stats in R
By Safa Andac
