Density Area

Description

A convenience function to get just the areas of density polygons.

Usage

density_area(
  x,
  y,
  probs = 0.5,
  as_sf = FALSE,
  as_list = FALSE,
  range_mult = 0.25,
  rangex = NULL,
  rangey = NULL,
  ...
)

Arguments

Details

If both rangex and rangey are defined, range_mult will be disregarded. If only one or the other of rangex and rangey are defined, range_mult will be used to produce the range of the undefined one.

Value

A list of data frames, if as_list=TRUE, or just a data frame, if as_list=FALSE.

Data frame output

If as_sf=FALSE, the data frame has the following columns:

level_id

An integer id for each probability level

prob

The probability level (originally passed to probs)

area

The area of the HDR polygon

sf output

If as_sf=TRUE, the data frame has the following columns:

level_id

An integer id for each probability level

prob

The probability level (originally passed to probs)

geometry

The sf::st_polygon() of the HDR

area

The area of the HDR polygon

Examples

library(densityarea)
library(dplyr)
library(sf)

ggplot2_inst <- require(ggplot2)

# basic usage

set.seed(10)
x <- rnorm(100)
y <- rnorm(100)

density_area(x,
             y,
             probs = ppoints(50)) ->
  poly_areas_df

head(poly_areas_df)

# Plotting the relationship between probability level and area
if(ggplot2_inst){
  ggplot(poly_areas_df,
         aes(prob, area)) +
    geom_line()
}

# Tidyverse usage

data(s01)

## Data preprocessing

s01 |>
  mutate(log_F2 = -log(F2),
         log_F1 = -log(F1)) ->
  s01

### Data frame output

s01 |>
  group_by(name) |>
  reframe(density_area(log_F2,
                       log_F1,
                       probs = ppoints(10))) ->
  s01_areas_df

if(ggplot2_inst){
  s01_areas_df |>
    ggplot(aes(prob, area)) +
    geom_line()
}

### Including sf output

s01 |>
  group_by(name) |>
  reframe(density_area(log_F2,
                       log_F1,
                       probs = ppoints(10),
                       as_sf = TRUE)) |>
  st_sf() ->
  s01_areas_sf

if(ggplot2_inst){
  s01_areas_sf |>
    arrange(desc(prob)) |>
    ggplot() +
    geom_sf(aes(fill = area))
}

Density polygons

Description

Given numeric vectors x and y, density_polygons() will return a data frame, or list of a data frames, of the polygon defining 2d kernel densities.

Usage

density_polygons(
  x,
  y,
  probs = 0.5,
  as_sf = FALSE,
  as_list = FALSE,
  range_mult = 0.25,
  rangex = NULL,
  rangey = NULL,
  ...
)

Arguments

Details

When using density_polygons() together with dplyr::summarise(), as_list should be TRUE.

If both rangex and rangey are defined, range_mult will be disregarded. If only one or the other of rangex and rangey are defined, range_mult will be used to produce the range of the undefined one.

Value

A list of data frames, if as_list=TRUE, or just a data frame, if as_list=FALSE.

Data frame output

If as_sf=FALSE, the data frame has the following columns:

level_id

An integer id for each probability level

id

An integer id for each sub-polygon within a probabilty level

prob

The probability level (originally passed to probs)

x, y

The values along the original x and y dimensions defining the density polygon. These will be renamed to the original input variable names.

order

The original plotting order of the polygon points, for convenience.

sf output

If as_sf=TRUE, the data frame has the following columns:

level_id

An integer id for each probability level

prob

The probability level (originally passed to probs)

geometry

A column of sf::st_polygon()s.

This output will need to be passed to sf::st_sf() to utilize many of the features of sf.

Examples

library(densityarea)
library(dplyr)
library(purrr)
library(sf)

ggplot2_inst <- require(ggplot2)
tidyr_inst <- require(tidyr)

set.seed(10)
x <- c(rnorm(100))
y <- c(rnorm(100))

# ordinary data frame output
poly_df <- density_polygons(x,
                            y,
                            probs = ppoints(5))

head(poly_df)

# It's necessary to specify a grouping factor that combines `level_id` and `id`
# for cases of multimodal density distributions
if(ggplot2_inst){
  ggplot(poly_df, aes(x, y)) +
    geom_path(aes(group = paste0(level_id, id),
                  color = prob))
}

# sf output
poly_sf <- density_polygons(x,
                            y,
                            probs = ppoints(5),
                            as_sf = TRUE)

head(poly_sf)

# `geom_sf()` is from the `{sf}` package.
if(ggplot2_inst){
  poly_sf |>
    arrange(desc(prob)) |>
    ggplot() +
    geom_sf(aes(fill = prob))
}

# Tidyverse usage

data(s01)

# Data transformation
s01 <- s01 |>
  mutate(log_F1 = -log(F1),
         log_F2 = -log(F2))

## Basic usage with `dplyr::reframe()`
### Data frame output
s01 |>
  group_by(name) |>
  reframe(density_polygons(log_F2,
                           log_F1,
                           probs = ppoints(5))) ->
  speaker_poly_df

if(ggplot2_inst){
  speaker_poly_df |>
    ggplot(aes(log_F2, log_F1)) +
    geom_path(aes(group = paste0(level_id, id),
                  color = prob)) +
    coord_fixed()
}

### sf output
s01 |>
  group_by(name) |>
  reframe(density_polygons(log_F2,
                           log_F1,
                           probs = ppoints(5),
                           as_sf = TRUE)) |>
  st_sf() ->
  speaker_poly_sf

if(ggplot2_inst){
  speaker_poly_sf |>
    ggplot() +
    geom_sf(aes(color = prob),
            fill = NA)
}

## basic usage with dplyr::summarise()
### data frame output

if(tidyr_inst){
  s01 |>
    group_by(name) |>
    summarise(poly = density_polygons(log_F2,
                                      log_F1,
                                      probs = ppoints(5),
                                      as_list = TRUE)) |>
    unnest(poly) ->
    speaker_poly_df
}
### sf output

if(tidyr_inst){
  s01 |>
    group_by(name) |>
    summarise(poly = density_polygons(
      log_F2,
      log_F1,
      probs = ppoints(5),
      as_list = TRUE,
      as_sf = TRUE
    )) |>
    unnest(poly) |>
    st_sf() ->
    speaker_poly_sf
}

Vowel Space Data

Description

This is the vowel space data from a single speaker, s01, whose audio interview and transcription are part of the Buckeye Corpus (Pitt et al. 2007). The transcript was realigned to the audio using the Montreal Forced Aligner (McAullife et al. 2022) and vowel formant data extracted with FAVE (Rosenfelder et al. 2022).

Usage

s01

Format

s01

A dataframe with 4,245 rows and 10 columns

name

Speaker id

age

Speaker age (y=young, o=old)

sex

Speaker sex

word

Word in the transcription

vowel

Arpabet transcription of the measured vowel

plt_vclass

A modified Labov/Trager notation of the measured vowel

ip_vclass

An IPA-like transcription of the measured vowel

F1

The measured F1 frequency (Hz)

F2

The measured F2 frequency (Hz)

dur

The measured vowel duration

Source

McAuliffe, M., Fatchurrahman, M. R., GalaxieT, NTT123, Amogh Gulati, Coles, A., Veaux, C., Eren, E., Mishra, H., Paweł Potrykus, Jung, S., Sereda, T., Mestrou, T., Michaelasocolof, & Vannawillerton. (2022). MontrealCorpusTools/Montreal-Forced-Aligner: Version 2.0.1 (v2.0.1) doi:10.5281/ZENODO.6658586

Pitt, M. A., Dilley, L., Johnson, K., Kiesling, S., Raymond, W., Hume, E., & Fosler-Lussier, E. (2007). Buckeye Corpus of Conversational Speech (2nd release). Department of Psychology, Ohio State University. https://buckeyecorpus.osu.edu/

Rosenfelder, I., Fruehwald, J., Brickhouse, C., Evanini, K., Seyfarth, S., Gorman, K., Prichard, H., & Yuan, J. (2022). FAVE (Forced Alignment and Vowel Extraction) Program Suite v2.0.0 https://github.com/JoFrhwld/FAVE