Per-patient trajectories across timepoints, split by outcome

WHY LINES AND NOT A BOX PER TIMEPOINT. In a repeated-measures cohort the signal is usually the DIRECTION each patient moves, and a box per timepoint averages exactly that away: two patients rising and two falling look like a flat cohort. In the sepsis literature this is the figure that separates survivors from non-survivors, because they differ in trajectory rather than at any one timepoint.

Usage

fig_clinical_trajectory(
  freq,
  time_of,
  patient_of,
  outfile,
  outcome_of = NULL,
  outcome_name = "outcome",
  value_col = NULL,
  ncol = 4L,
  time_levels = NULL,
  dpi = 200,
  colors = fcs_colors()
)

Arguments

freq

population_frequencies table.

time_of

named vector sample_id -> timepoint.

patient_of

named vector sample_id -> patient.

outfile

path.

outcome_of

optional named vector sample_id -> outcome, used to colour the lines and to split the mean.

outcome_name

label for the outcome in the key.

value_col

frequency column.

ncol

panels per row.

time_levels

order of the timepoints; sorted unique values by default.

dpi

resolution.

colors

palette.

Details

WHAT IT IS NOT. Descriptive. The test for a repeated-measures design is the paired comparison in paired_comparison_stats.csv; a trajectory figure shows the shape of the data that test was run on.