Toolbox and Skillset of Reproducible Research
Andrey Koval
October 14, 2014
bio
Workflow in research projects
Academic Paper:
- Introduction
- Methods
- Results
- Discussion
Workflow in research projects
Workflow in research projects
Workflow in research projects
Advantages of RR
- Less prone to human error
- Removes tedious jobs
- Recordable workflow
- Correcting mistakes later
- Synchronizes analysis and discussion
TOOLBOX and toolset of RR
- R
- RStudio
- GitHub
TOOLBOX : R
TOOLBOX : RStudio
TOOLBOX : GitHub
toolbox and SKILLSET of RR
- Data Manipulation
- Graph Production
- Statistical Modeling
Dynamic Reporting
see ialsa.github.io/tutorials for knowledge base
Goals of RR
Ultimate : Answering a research question
Practical : Publishing a paper, producing a manuscript
Technical : Producing a dynamic document
RR uses code to
In the process of achieving these goals we encounter the need to perform particular tasks on a computer:
A. Load and inspect datasets
B. Depict data in statistical graphics
C. Fit statistical models
D. Compare observed and modeled data
E. Produce reports discussing the steps above
let's demonstrate how we can accomplish each of these tasks on a R simulator
A. Load and inspect datasets
B. Depict data as statistical graphics
require(ggplot2)
p <- ggplot(mtcars, aes(x=wt, y=mpg))
p <- p + geom_point()
p
C. Fit statistical models
D. Compare observed and modeled data
require(ggplot2)
ds <- mtcars
ds$mpg_modeled <- predict(lm(mpg ~ wt, mtcars))
p <- ggplot(ds, aes(x=wt, y=mpg))
# p <- p + geom_point()
p <- p + geom_point(aes(y=mpg_modeled),color="red")
p
D. Compare observed and modeled data
require(ggplot2)
ds <- mtcars
ds$mpg_modeled <- predict(lm(mpg ~ wt, mtcars))
p <- ggplot(ds, aes(x=wt, y=mpg))
p <- p + geom_point()
p <- p + geom_line(aes(y=mpg_modeled),color="red")
p
E. Produce report
Recap
Traditional
vs Reproducible