3 Introduction

rtemis is under active development. New algorithms will be added to the public repository as their corresponding papers are published. The package has already been used to tune, train, and test a very large number of experiments.
This website is also a work in progress - more chapters will be added and existing content will be expanded.

3.1 Powerful and efficient data science

rtemis provides a common interface to a large number of algorithms for supervised learning, decomposition, and clustering.
The goal is to make data science as efficient and as painless as possible while maintaining full control of model training. The technical demands of data analysis should not come between a scientist and their data.

3.2 Low level to high level functions

rtemis provides both low- and high-levels functions. It allows you to perform many operations with a single command: train a single model, tune and train a model (internal cross-validation), train and test a model by crossvalidation (external crossvalidation), tune and test a model (internal + external crossvalidation aka nested resampling).

3.3 R6 classes

rtemis makes extensive use of R6 classes.
R6 objects contain both attributes and functions - you have everything you need in one place. The different chapters in this book will show you how convenient they are to use. R6 classes also support S3 methods, so you have access to all the familiar R generics as well.

3.4 Naming conventions

rtemis uses the following naming conventions for supervised models:

  • fitted values refer to the estimated outcome of the training data - the values that were actually fitted by the model
  • predicted values refer to the estimated outcome of the testing data
  • estimated values refer to fitted and predicted values collectively.

fited and predicted fields appear within supervised model objects themselves, think of them as y-hat train and y-hat test respectively.
estimated appears in function arguments, for example classError(true, estimated) since they could refer to either fitted or predicted values.

3.5 Citing rtemis

citation("rtemis")

To cite rtemis in publications, please use:

  Gennatas, E. D. (2017). Towards Precision Psychiatry: Gray Matter
  Development and Cognition in Adolescence (Doctoral dissertation,
  University of Pennsylvania).

A BibTeX entry for LaTeX users is

  @PhdThesis{,
    title = {Towards Precision Psychiatry: Gray Matter Development and Cognition in Adolescence},
    author = {{Gennatas} and Efstathios D.},
    year = {2017},
    school = {University of Pennsylvania},
  }