Customising ProjectTemplate in R

ProjectTemplate is an R Bundle which facilitates information evaluation, encourages good information evaluation habits, and standardises many information analytic steps. After a few years of refining an information evaluation workflow in R, I realised that I might mainly converged on one thing just like ProjectTemplate anyway. Nonetheless, my method was not fairly as systematic, and it took extra effort than essential to get began on a brand new mission. Thus, since late 2013, I have been utilizing ProjectTemplate to organise my R information evaluation tasks.
Whereas I’ve discovered ProjectTemplate to be a wonderful instrument, I realised that once I created a brand new information evaluation mission based mostly on ProjectTemplate, I used to be repeatedly making numerous customisations to the preliminary set of information and folders. Thus, I’ve now arrange a repository to retailer these customisations in order that I can get began on a brand new information evaluation mission extra effectively. The aim of this put up is to doc these modifications.
This put up assumes an affordable information of R and ProjectTemplate. For those who’re not acquainted with ProjectTemplate, you might take a look at the ProjectTemplate web site focusing significantly on the Getting Began part. For those who’re actually eager you might additionally watch an hour lengthy video on ProjectTemplate, RStudio, and GitHub

Common setup

I’ve a duplicate of my customised model of the ProjectTemplate listing and file construction on github within the AnglimModifiedProjectTemplate repository. Particularly, it has:

1. Modifications to world.dcf as described beneath,
2. a clean readme.md
3. a few directories eliminated that I do not use (e.g., diagnositics, logs, profiling)
4. an preliminary rmd file with the customisations talked about beneath within the studies listing
5. An .Rproj RStudio mission file to allow simple launching of RStudio.
6. A further output listing for storing tabular, textual content, and different output

Thus, at any time when I need to begin a brand new information evaluation mission I can obtain and extract the zip file of the repository on github).
Thus, after making a mission folder, the next steps could be skipped when utilizing my customised template.

• Open RStudio and create RStudio Challenge in current listing
• Create ProjectTemplate folder construction with library(ProjectTemplate); create.mission()
• Transfer ProjectTemplate information into folder
• Modify world.dcf
• Setup rmd studies

I additionally doc beneath a couple of further factors about subsequent steps together with:

• Organising the information listing
• Updating the readme file
• Setttig up git repository

Modifying world.dcf

My most well-liked beginning world.dcf settings are

data_loading: on
cache_loading: off
munging: on
logging: off
load_libraries: on
libraries: psych, lattice, Hmisc
as_factors: off
data_tables: off

A bit of clarification:

• as_factors I do fairly a little bit of string processing, significantly on meta information and on output tables. I discover the automated conversion of strings into components to be a extremely annoying characteristic. Thus, setting this to off is my most well-liked setting.
• load_libraries: I all the time have further libraries so it is sensible to have this on.
• libraries: There are a lot of frequent packages that I exploit, however I virtually all the time make use of the above comma separate listing of packages.

Setup rmd information

Fundamentals of such information

The primary line within the first chunk is all the time:

```{r}
library(ProjectTemplate); load.mission()
```

This hundreds the whole lot required to get began with the mission.

Setup information folder

ProjectTemplate mechanically names ensuing information.frames with a reputation based mostly on the file title. That is handy. Nonetheless, it’s usually the case that the file names have to be modified from some uncooked information equipped or it could be that the unique information format shouldn’t be completely fitted to importing. In that case, I retailer the uncooked information in a separate folder referred to as raw-data after which export or create a duplicate within the desired format with the specified title within the information folder.

Overriding default information import choices

Some information information cannot be imported utilizing the default information import guidelines. After all, you may change the file to adjust to the principles. Alternatively, I believe the usual answer is so as to add a file within the lib listing (e.g., data-override.r) that imports the information information. Give the imported information file the identical title that ProjectTemplate would.

Replace readme

I modify the file to README.md to make it clear that it’s a markdown formatted file. I can then add slightly details about the mission.

Setup git repository

If utilizing github, I create a brand new repository on github.

Output folder

A typical workflow for me is to generate tables, textual content, and determine output fromthe script which is then integrated right into a manuscript doc. Whereas I actually like Sweave and RMarkdown, I usually discover it extra sensible to write down a manuscript in Microsoft Phrase. I exploit the output folder to retailer tabular output, normal textual content output, and figures.
Within the case of tabular output, there may be the duty of making certain the desk is formatted appropriately (e.g., desired variety of decimal locations, cell alignment, cell borders, font, cell merging, and so forth.). I sometimes discover this best to do in Excel. Thus, I’ve a file referred to as output-processing.xlsx. I import the tabular information into this file and apply related formatting. This may then be integrated into the manuscript. Listed here are a couple of extra notes about Desk conversion in MS Phrase.