Skip to contents

Plot the fit index of a grid search

Source: R/plotGridSearch_gg.R
plotGridSearch_gg.Rd

Draws the fit index recorded by exametrika::GridSearch() against the number of classes or ranks. For LCA/LRA this is one line. For Biclustering the default is one line per number of fields, and style = "tile" draws the class-by-field grid as a heat map instead. Settings that converged are filled points, settings that left a class, rank or field empty are open triangles, settings that failed to converge are left blank, and the optimum is marked with a large red point (and, for one line, a dotted vertical line).

Usage

plotGridSearch_gg(
  data,
  style = c("line", "tile"),
  title = TRUE,
  colors = NULL,
  linetype = "solid",
  show_legend = TRUE,
  legend_position = "right"
)

Arguments

data

A GridSearch result from exametrika::GridSearch().

style

Character. "line" (default) draws the index against the number of classes/ranks; "tile" (Biclustering only) draws a heat map over classes and fields.

title

Logical or character. If TRUE (default), display an auto-generated title. If FALSE, no title. If a character string, use it as a custom title.

colors

Character vector. For style = "line": the line color (LCA/LRA) or one color per number of fields (Biclustering). For style = "tile": two colors for the low and high ends of the index scale. If NULL (default), the package palette is used.

linetype

Character or numeric. Line type of the index line(s). Default is "solid". Ignored for style = "tile".

show_legend

Logical. If TRUE (default), display the legend.

legend_position

Character. Position of the legend. One of "right" (default), "top", "bottom", "left", "none".

Value

A single ggplot object.

Details

The plot reads index_vec (LCA/LRA) or index_matrix (Biclustering) together with the empty_vec/empty_matrix flags that GridSearch() records since exametrika 2.3.0. With an older exametrika no setting is marked as empty.

Examples

# \donttest{
library(exametrika)
gs <- GridSearch(J15S500, max_ncls = 5, fun = "LCA", index = "BIC")
#> 
#> Optimal ncls/nrank is 5
#> Running analysis with optimal parameters...
plotGridSearch_gg(gs)


gs_b <- GridSearch(J15S500, max_ncls = 4, max_nfld = 4,
  fun = "Biclustering", index = "BIC")
#> 
#> Optimal ncls/nrank is  4 and Optimal nfld is 4
#> Running analysis with optimal parameters...
plotGridSearch_gg(gs_b)

plotGridSearch_gg(gs_b, style = "tile")

# }