In joekroese/tidymine: Tidy interface for calculating Maximal Information-based Nonparametric Exploration (MINE) statistics

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(tidyverse)
library(tidymine)

Motivation

We often want to work out which variables are most relevant for predicting or describing another.

The Limitations of Correlation

Correlation struggles with this because it is mainly limited to recognising linear patterns.

Here are some examples.

MINE Statistics

MINE statistics get around this problem.

mlb_2008 %>%
  tidymine(salary) %>%
  arrange(-mic)

joekroese/tidymine documentation built on Aug. 25, 2020, 12:06 a.m.