README.md

In erblast/oetteR: Collection of personal R functions

oetteR

This is my personal R package which contains a number of functions that help me to maintain an organized workflow. Most of the functions are just curried versions of functions from other packages, nevertheless I added a decent amount of documentation, however there are a few packages that I cosntantly use and that one should know about in order to understand the documentation.

Prerequisites

• Pipes and the Tidyverse
• Functional dataframe based modelling concepts as described in the R4DS book linked above. A package that I started out with is pipelearner which provides an easy to use interface for modelling dataframes and I use it in some of my functions. However, pipelearner always stores the entire model inside the modelling df and you quickly run into memory issue so I tend to write my own modelling dataframes and to use the tools devloped by Max Kuhn rsample recipes and caret.
• I generate a lot of my visual results as html and use rmarkdown to its full capacity. In my project folder I usually have one folder for Rmd and one for html files. I have one Rmd file for each step in my workflow and all resulting hmtls are rendered to the html folder. I have one execute.R file in the parent project folder which triggers rendering of all Rmds. The last Rmd file to be rendered will generate a index.html file which is basically a catalogue file with links to all htmls. This index.html file can be found in the project parent folder. I try to use widgets over static plots and tables whenever I can. It helps to know the following packages at least bit webshot, bookddown, htmltools, knitr, plotly, DT

Naming

The names of all functions start with f_ followed by another prefix that describes the role of the function inside my workflow

Workflow (including links to vignettes)

Variable Manipulations and Transformations f_manip*

Data Preparation f_clean_*

When thinking about this step I did not know about the recipes package. I came up with something similar but less elabprate and thought-through.

• Data Perparation and PCA

Data Exploration and Visualisation f_pca_*, f_stat_*, f_plot_*

• Alluvial Plots
• Group Analysis

Feature Testing f_model_importance_*

• Leakage and variable importance for modelling

Model Training f_train_*

• Gamma/Tweedie Regression with Lasso

Model Visualisation f_model_plot_*

• Visualising Regression Models

Reporting f_html_*, f_datatable, f_plot_obj_2_html

These functions help me to generate html output.

Apps f_shiny_*, f_simulation_*

These functions start a shinyserver and let you run a simulation or a shiny app.

• f_shiny_multiview runs a tool for analysing labeled groups in a data set
• f_shiny_som runs a tool for training self organizing maps. The subsequent clustering algorithm allows you to cluster only adjacent areas of the map.

You can pass your own data via the data argument to both functions, or select from a variety of sample datasets if data == NULL.

Additional Content

Apart from the vignettes that I have linked to already above I have added some files that I produced as prrof of concept (POC) files. When I try out a new package or get stuck somewhere with a problem I like to write those as a minimal example of how to solve or not to solve a problem. One can use the function f_content() to produce and open a index html file with links to all those files.

Some of the POCs are also posted on Rpubs

Visualisations

• clickable file links in DT::datatable
• mix taglist and plots
• ggalluvial
• ggpubr

Modelling

Modelling Dataframe packages

• Pipelearner for classification
• rsample
• recipes
• caret

Regression models with caret

• 01 robust linear regression, rlm
• 02 neuronal networks, nnet
• 03 multiviariate adaptive regession splines (MARS), earth

Various

• WOE, IV Scorecard in Credit Risk Modelling

Algorithms

• SOM with connectivity restrains
• Prediction Intervalls

erblast/oetteR documentation built on May 27, 2019, 12:11 p.m.