Parallel Coordinates

A parallel coordinates plot places multiple variables on parallel vertical axes and draws a line for each observation connecting its values across all axes. It is the most effective way to visualize multi-variable profiles —showing which variables co-vary and which distinguish groups —without requiring dimensionality reduction.

STEP 1 —When to Use

Use parallel coordinates when: you have 4—5 continuous variables measured on the same subjects; you want to show the full multi-variable profile of each group (or each subject); you want to identify which variables most clearly separate groups; you are comparing group profiles rather than individual variable means.

Do not use when: variables are on very different scales without normalization (lines will cluster near the extremes); you have too many subjects/lines (n > 50 creates unreadable overplotting —use group means or medians instead of individual lines); variables are categorical (categorical parallel coordinates require a different approach).

[Image placeholder: Parallel coordinates plot with 6 variables (axes) and 2 groups (30 lines each, colored red vs. blue, 50% transparency). Group A lines trend upward on axes 3—; Group B lines are flat. Mean profile per group overlaid as thick lines. The visual separation pattern between groups is clearly visible on specific axes.]

Parallel coordinates: individual lines (transparent) show the full profile variability; thick mean profiles show the group separation pattern across variables.

STEP 2 —Journal Requirement

Axis normalization method must be stated: min-max, z-score, or none (raw values on separate scales).
Whether individual lines or group means/medians are displayed must be stated.
n per group and missing data handling must be disclosed.
Color palette for groups must be stated (and must be colorblind-accessible).

STEP 3 —Licklider Implementation

Input

Numeric matrix with multiple variable columns
Categorical grouping variable (color)
Axis normalization: min-max (default), z-score, or raw
Display mode: individual lines (n < 50 per group) or group summaries (mean ﾂｱ SD or median + IQR)

Output

Parallel axes with normalized scales, original scale tick labels
Individual lines with group color and transparency (alpha = 0.3—.6 based on n)
Group mean/median overlaid as thick lines
Variable axis labels with units
Missing data: lines interrupted at missing values (not connected across gaps)

[Image placeholder: Licklider parallel coordinates with 7 variables (z-score normalized), 2 groups (n=20 each). Individual lines at 30% transparency, group mean lines at full opacity (thick). Axes labeled with original units and scale marks. One variable has 2 missing values (lines interrupted). Legend: "Lines: individual observations (alpha=0.3). Thick lines: group means. Axes: z-scored."]

Licklider parallel coordinates: z-score axes with original scale marks, individual + mean lines, interrupted lines for missing values, legend disclosing normalization and n.

STEP 4 —Draft Output (Draft / Needs review)

Parallel coordinates plot of [N variables] for [M groups]
(n = 20 per group). Axes are z-score normalized (axis tick marks show
original scale values). Individual observations are shown as thin lines
(alpha = 0.3). Group means are overlaid as thick lines. Two observations
with missing values for [Variable 4] are shown with interrupted lines.
No observations were excluded.