In schloerke/autocog: Automatic Cognostic Summaries

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = file.path("man", "figures", ""),
  fig.height = 8,
  fig.width = 12
)

autocogs

Cognostics are univariate statistics (or metrics) for a subset of data. When paired with the underlying data of visualizations, cognostics are a powerful tool for ordering and filtering the visualizations. add_panel_cogs() will automatically append cognostics for each plot player in a given panel column. The newly appended data can be fed into a trelliscopejs widget for easy viewing.

Installation

You can install autocogs from github with:

remotes::install_github("schloerke/autocogs")

Examples

Gapminder

library(autocogs)
library(tidyverse)
library(gapminder)
# remotes::install_github("hafen/trelliscopejs")
# remotes::install_github("schloerke/trelliscopejs@autocogs")
library(trelliscopejs)

# Explore
p <-
  ggplot(gapminder, aes(year, lifeExp)) +
  geom_line(aes(group = country)) +
  geom_smooth(method = "lm", formula = y ~ x)
p

Looking at the plot above, most countries follow a linear trend: As the year increases, life expectancy goes up. A few countries do not follow a linear trend.

In the examples below, we will extract cognostics to aid in exploring the countries whose life expectancy is not linear.

`trelliscopejs::facet_trelliscope()`

ggplot(gapminder, aes(year, lifeExp)) +
  geom_smooth(method = "lm", formula = y ~ x) +
  geom_line() +
  trelliscopejs::facet_trelliscope(
    ~ country + continent,
    nrow = 3, ncol = 6,
    self_contained = TRUE,
    auto_cog = TRUE,
    state = list(
      # set the state to display the country, continent, and R^2 value
      #   sorted by ascending R^2 value
      sort = list(trelliscopejs::sort_spec("_lm_r2")),
      labels = c("country", "continent", "_lm_r2")
    ),
    path = "readme-figs/facet"
  )
# (screen shot of trelliscopejs widget)

`trelliscopejs::trelliscope()`

This is a full, start to finish example how automatic cognostics could be inserted into a data exploration workflow.

# Find a consistent y range
y_range <- range(gapminder$lifeExp)

## # Set up data and panel column
gapminder %>%
  group_by(country, continent) %>%
  # nest the data according to the country and continent
  nest() %>%
  mutate(
    # create a column of plots with a
    # * line
    # * linear model
    panel = lapply(data, function(dt) {
      ggplot(dt, aes(year, lifeExp)) +
        geom_smooth(method = "lm", formula = y ~ x) +
        geom_line() +
        ylim(y_range[1], y_range[2])
    })
  ) %>%
  print() ->
gap_data

# Double check the plot worked...
# Look at the first panel (ggplot2 plot) of Afghanistan
gap_data$panel[[1]]

#!!!!!!!!!!
# Add cognostic information given the panel column plots
#!!!!!!!!!!
gap_data %>%
  autocogs::add_panel_cogs() %>%
  ungroup() %>%
  # double check it was added
  print(width = 100) ->
full_gap_data

# Display the panel and cognostics in a trelliscopejs widget
trelliscopejs::trelliscope(
  full_gap_data, "gapminder life expectancy",
  panel_col = "panel",
  ncol = 6, nrow = 3,
  auto_cog = FALSE,
  self_contained = TRUE,
  state = list(
    # sort by ascending R^2 value (percent explained by linear model)
    sort = list(trelliscopejs::sort_spec("_lm_r2")),
    # display the country, continent, and R^2 value
    labels = c("country", "continent", "_lm_r2")
  ),
  path = "readme-figs/manually"
)
# (screen shot of trelliscopejs widget)

Custom Cognostics

add_cog_group() to add a custom cognostics group.
add_layer_cogs() to call which cognostics groups should be executed for a given plot layer.

Using existing code from the autocogs package, we will add the univariate continuous cognostics group.

add_cog_group(
  "univariate_continuous",
  field_info("x", "continuous"),
  "univariate metrics for continuous data",
  function(x, ...) {
    x_range <- range(x, na.rm = TRUE)
    list(
      min = cog_desc(x_range[1], "minimum of non NA data"),
      max = cog_desc(x_range[2], "maximum of non NA data"),
      mean = cog_desc(mean(x, na.rm = TRUE), "mean of non NA data"),
      median = cog_desc(median(x, na.rm = TRUE), "median of non NA data"),
      var = cog_desc(var(x, na.rm = TRUE), "variance of non NA data")
    )
  }
)

We can then call the 'univariate_continuous' cognostics group whenever a geom_rug layer is added in a ggplot2 plot object using the code below.

add_layer_cogs(
  # load_all(); p <- qplot(x = 1, y = Sepal.Length, data = iris, geom = "boxplot"); plot_cogs(p)
  "geom_boxplot",
  "boxplot plot",
  cog_group_df(
    "univariate_continuous", "y", "_y",
    "boxplot", "y", "_boxplot",
    "univariate_counts", "y", "_n"
  )
)