Skip to contents

Figures of a principal component analysis

Source: R/plot-pca.R
plot_pca_scree.Rd

  • plot_pca_scree(): variance explained per component and cumulative variance, with the Kaiser line (100 / p %).

  • plot_pca_ind(): individuals map, coloured by a variable of data (90% normal-theory ellipses for groups).

  • plot_pca_var(): correlation circle.

  • plot_pca_contrib(): contribution of each variable to each component, 100 * loading^2 (sums to 100% per component), with the expected contribution if all variables contributed equally.

  • plot_pca_distribution(): distribution of the component scores by the levels of categorical variables, with plot_distribution().

Usage

plot_pca_scree(pca_obj, n_pc = 10, base_size = 11, title = "Scree plot")

plot_pca_ind(
  pca_obj,
  data = NULL,
  id_col = NULL,
  color_var = NULL,
  labels = NULL,
  axes = c(1, 2),
  palette = NULL,
  ellipse = TRUE,
  base_size = 11,
  title = "PCA: individuals"
)

plot_pca_var(
  pca_obj,
  labels = NULL,
  axes = c(1, 2),
  base_size = 11,
  groups = NULL,
  title = "PCA: correlation circle"
)

plot_pca_contrib(
  pca_obj,
  axes = 1:2,
  labels = NULL,
  groups = NULL,
  palette = NULL,
  top_n = NULL,
  base_size = 11,
  title = "PCA: variable contributions"
)

plot_pca_distribution(
  pca_obj,
  data,
  group_vars,
  id_col = NULL,
  pcs = 1:2,
  labels = NULL,
  palette = NULL,
  base_size = 10,
  ...
)

Arguments

pca_obj

Output of pca_compute().

n_pc

Number of components shown.

base_size

Base font size.

title

Title.

data

Data frame holding the colour / grouping variables.

id_col

Identifier column used in pca_compute().

color_var

Colour variable (name or bare symbol).

labels

Optional named vector of labels.

axes

Components shown (two for the maps, any number for the contributions).

palette

Colours of the groups.

ellipse

Draw 90% ellipses per group.

groups

Optional named list group -> variables (colour of the arrows or bars).

top_n

Keep the top_n contributions per component.

group_vars

Categorical variables (names or tidyselect).

pcs

Components shown.

...

Passed to plot_distribution().

Value

A ggplot; plot_pca_distribution() returns a named list of patchworks, one per grouping variable.

Examples

pc <- pca_compute(cohort_sim, c(NFL_V0, IL6_V0, CRP_V0, MOCA_V0, UPDRS_V0, AGE), id_col = ID)
plot_pca_scree(pc)

plot_pca_ind(pc, cohort_sim, id_col = ID, color_var = GROUP)

plot_pca_var(pc, groups = list(Biomarkers = c("NFL_V0", "IL6_V0", "CRP_V0")))

plot_pca_contrib(pc, axes = 1:2)

plot_pca_distribution(pc, cohort_sim, GROUP, id_col = ID)[[1]]