Paired visits: one line per subject between two visits

Two overlaid distributions (V0 and V3) can be identical while every subject moves. This plot shows the subject-level changes: one line per subject, a boxplot per visit and the median line. The panel title gives the number of paired subjects, the Spearman correlation between visits (stability of the ranking) and the paired Wilcoxon test (shift of the distribution).

Usage

plot_paired_visits(
  data,
  vars,
  id_col,
  visits = c("V0", "V3"),
  labels = NULL,
  ncol = 4,
  palette = c("#0072B2", "#D55E00"),
  title = NULL,
  subtitle = NULL,
  base_size = 10,
  transitions = FALSE,
  split = c("median", "mean", "quantile", "manual"),
  split_prob = 0.5,
  split_cut = NULL,
  split_ref = c("visit", "first"),
  transition_mode = c("detailed", "crossing"),
  transition_palette = NULL
)

Arguments

data

One row per subject (wide format).

vars

First-visit numeric variables (names or tidyselect), e.g. NFL_V0; the second-visit column is found with visit_swap().

id_col

Subject identifier.

visits

The two visit suffixes.

labels

Optional named vector of labels.

ncol

Number of columns of panels.

palette

Colours of the two visits.

title, subtitle

Titles.

base_size

Base font size.

transitions

Colour the lines by High / Low transition.

split

Threshold rule: "median", "mean", "quantile" (split_prob) or "manual" (split_cut, then required for every variable).

split_prob

Probability of the "quantile" rule.

split_cut

Manual threshold(s), named by first-visit variable (override split for these variables), or one unnamed value for all, on the plotted scale.

split_ref

"visit": threshold computed at each visit (insensitive to a global shift); "first": threshold of the first visit applied at both. Manual thresholds are the same at both visits.

transition_mode

"detailed" (High -> High, Low -> Low, High -> Low, Low -> High) or "crossing" (Same side, High -> Low, Low -> High).

transition_palette

Named colours of the transitions (NULL: default).

Value

A ggplot with the attributes "transitions" (one row per subject x variable, when transitions = TRUE) and "summary" (one row per variable: n, rho, Wilcoxon p, thresholds, % changing side, kappa).

Details

With transitions = TRUE, each subject is "High" or "Low" at each visit with respect to a threshold (median by default) and its line is coloured by its transition; the threshold of each visit is drawn as a dashed segment, and the panel title adds the share of subjects changing side and Cohen's kappa of the High / Low status (kappa < 0: inverted ranking).

Examples

plot_paired_visits(cohort_sim, c(NFL_V0, MOCA_V0), id_col = ID)

p <- plot_paired_visits(cohort_sim, NFL_V0, id_col = ID, transitions = TRUE)
attr(p, "summary")
#> # A tibble: 1 × 8
#>   var        n   rho        p    t0    t1 pct_change kappa
#>   <chr>  <int> <dbl>    <dbl> <dbl> <dbl>      <dbl> <dbl>
#> 1 NFL_V0   107 0.964 9.21e-10  12.4  14.8       5.61 0.888