Create a Taylor diagram to assess model performance using the relationship between correlation, standard deviation, and centered RMS error.

Blah Blah Blah Taylor (2001) Blah Blah Blah TODO: Add description

Usage

taylor_diagram(
  dat,
  data_cols = c(obs = "obs", mod = "mod"),
  group_by,
  facet_by = NULL,
  facet_rows = 1,
  obs_colour = "purple",
  obs_shape = 16,
  obs_size = 1.5,
  obs_stroke = 1,
  obs_label = "Obs.",
  mod_colours = "default",
  mod_fills = "default",
  mod_shapes = "default",
  mod_size = 1.5,
  mod_stroke = 1,
  cor_minimum = NULL,
  cor_step = 0.1,
  cor_colour = "grey30",
  cor_linetype = "longdash",
  cor_label = "Correlation",
  rmse_minimum = 0,
  rmse_step = "default",
  rmse_colour = "brown",
  rmse_linetype = "dotted",
  rmse_label = "Centered RMS Error",
  rmse_label_pos = "default",
  sd_maximum = NULL,
  sd_step = "default",
  sd_colour = "black",
  sd_linetypes = c(obs = "dashed", other = "dashed"),
  sd_label = "Standard Deviation",
  plot_padding = 0.5,
  labels_padding = 2
)

Arguments

dat: Paired observation and model data with (at least) all columns in `data_cols`, `group_by`, and (if provided) `facet_by`.
data_cols: (Optional) a character vector with 2 values indication column names in `dat` to get observed and modelled values. Default assumes columns "obs" and "mod" exist.
group_by: a character vector with between 1 and 3 column names to use as groups. The first value will be used for `colour`, the second (if present) will be used for `shape`, and the third (if present) will be used for `fill` when adding model data points. If names are present they will be used as the corresponding legend titles.
facet_by: (Optional) a character vector with 1 or 2 column names to use as facets in `ggplot2::facet_wrap()`. If names are present they will be used as the corresponding facet titles. Default (NULL) does not facet the plot.
facet_rows: (Optional) a single numeric value indicating the number of rows to use in facetting if `facet_by` values provided. Default is a single row.
obs_colour, obs_shape, obs_size, obs_stroke: (Optional) a single value indicating the colour/shape/size/stroke of the observed data point. Default is a full-colour purple circle.
obs_label: (Optional) a single character value indicating the text to display for the observed point. Default is "Obs."
mod_colours, mod_shapes, mod_fills: (Optional) a named vector of colours/shapes to use for the provided `group_by` where the names correspond to values in that group column to assign each colour/shape to (i.e `c("group_1" = "red", ...)`). Default uses "good looking" colours/shapes/fills.
mod_size, mod_stroke: (Optional) a single numeric value indicating the size/stroke of the model data points. Default matches the size/stroke of the observed point.
cor_minimum: (Optional) a single numeric value indicating the minimum correlation value to display (from -1 to +1). Default uses the nearest 0.1 below the minimum correlation.
cor_step: (Optional) a single value indicating the spacing between each correlation line. Default is a step of 0.1 (10%).
cor_colour, cor_linetype: (Optional) a single value indicating the colour/linetype of the correlation grid lines. Default is grey long-dash lines.
cor_label: (Optional) a single character value indicating the text to display for the correlation axis title. Default is "Correlation".
rmse_minimum: (Optional) a single numeric value indicating the minimum rmse line to display (>= 0). Default is 0 (meaning the first line to display is at `rmse_step`).
rmse_step: (Optional) a single value indicating the spacing between each rmse line. Default produces approximatley 4 lines with "pretty" spacing.
rmse_colour, rmse_linetype: (Optional) a single value indicating the colour/linetype of the rmse circles originating from the observed point. Default is brown dotted lines.
rmse_label: (Optional) a single character value indicating the text to display for the RMSE axis title. Default is "Centered RMS Error".
rmse_label_pos: (Optional) a single value (0-1) indicating the location of the labels for the rmse circles (0 == far left along x-axis, 0.5 = top of cirles, 1 = far right along x-axis). Default is 10% greater than the minimum correlation.
sd_maximum: (Optional) a single numeric value indicating the maximum standard deviation value to display (>= 0). Default is the nearest "pretty" value above the maximum standard deviation.
sd_step: (Optional) a single value indicating the spacing between each standard deviation line. Default produces approximatley 4 lines with "pretty" spacing.
sd_colour: (Optional) a single value indicating the colour of the standard deviation arcs. Default is black.
sd_linetypes: (Optional) a character vector with 2 line types and names `"obs", "other"` indicating the line types of standard deviation arcs. Default is dashed for the observed line, dotted for others.
sd_label: (Optional) a single character value indicating the text to display for the standard deviation axis title. Default is "Standard Deviation".
plot_padding, labels_padding: (Optional) a single numeric value indicating how much spacing (standard deviation units) to add to most text labels. Default is 2 for both, likely needs to be adjusted depeding on the figure size and number of facets.

Value

A ggplot object of your taylor diagram.

Examples

if (FALSE) { # \dontrun{
# Make test data
data <- as.data.frame(datasets::ChickWeight) |> # TODO: make better test dataset
  dplyr::filter(.data$Chick == 1) |>
  tidyr::pivot_wider(names_from = "Chick", values_from = "weight") |>
  dplyr::full_join(
    as.data.frame(datasets::ChickWeight) |>
      dplyr::filter(.data$Chick != 1)
  ) |>
  dplyr::rename(obs = `1`, mod = "weight") |>
  dplyr::mutate(Chick = factor(round(as.numeric(.data$Chick) / 20)))
# Basic usage
taylor_diagram(data, group_by = c(Diet = "Diet", Chick = "Chick"))
# Force 0 on left axis
taylor_diagram(data,
  group_by = c(Diet = "Diet", Chick = "Chick"),
  cor_minimum = 0, rmse_label_pos = 130
) # TODO: fix this
# Change colours / shapes
taylor_diagram(data,
  group_by = c(Diet = "Diet", Chick = "Chick"),
  mod_colours = c("AB" = "pink", "BC" = "blue"),
  mod_fills = c("EGG" = "white", "PA" = "darkgrey"), # TODO: update this
  mod_shapes = c("FALSE" = 23, "TRUE" = 22),
  mod_size = 4, mod_stroke = 6,
  obs_colour = "brown", obs_shape = 23, obs_size = 6,
  cor_colour = "orange", cor_linetype = "dotdash",
  rmse_colour = "green", rmse_linetype = "longdash",
  sd_colour = "purple", sd_linetypes = c(obs = "solid", other = "dashed")
)
# Adjust text positioning
taylor_diagram(data,
  group_by = c(Diet = "Diet", Chick = "Chick"),
  plot_padding = 4, labels_padding = 1, rmse_label_pos = 0.7
)
} # }

Create a Taylor diagram to assess model performance using the relationship between correlation, standard deviation, and centered RMS error.

Usage

Arguments

Value

See also

Examples