VectorPaint scatterplots: map every point to two values

English translation of the Chinese original, translated on 2026-11-04. Original publication dates, attribution, code and version notes are retained.

Each scatterplot point represents a pair of values rather than its position in an observation sequence. Plot five fictional samples manually and check both axes. Unlike a sequential line chart, this exercise does not connect observations, automatically fit a relationship or establish that one variable causes the other.

Keep five value pairs and their meanings

List samples A–E outside the app with pairs (1,2), (2,5), (3,3), (4,6) and (5,4). Define the axes as fictional quantities A and B, both in relative teaching units, without real survey names. Every row needs a complete pair: if you sort x, its y partner must move with it. Create a 1200 by 800 pixel canvas and reserve space on all sides for axis names and labels. Record that there are only five samples and no time dimension; IDs serve solely to retrieve observations. Check the list before drawing coordinates so an attractive image cannot conceal broken pairings.

Build equal spacing separately on each axis

Use x from 0 to 6 and y from 0 to 8. Units must be equally spaced within each axis, although a unit need not occupy the same pixel length on both axes. Draw axes and ticks before their names and units so readers need not guess what numbers mean. Manually drawn references are artwork objects and export with it; check actual output for native grids or guides. Inspect the axes alone for origin, maxima and usable area. Never distort intervals to arrange the five points into a pleasing diagonal. Axis scaling affects apparent slope; a screen angle is not an uncalculated statistical conclusion.

Place identical markers at paired coordinates

Find A’s x value 1 and y value 2 and centre a small circle on their intersection, then place the remaining four points. Initially use identical size and fill so marker area does not imply an unprovided third variable. After each placement, verify ID and both values against the list: observation B has y=5, not height 2 merely because it is second. Read every coordinate before adding decorative lines. Available numerical positioning controls may help; otherwise establish relationships through grids and references without claiming unverified measurement precision from dragging. Keep every marker independently selectable.

Offset labels without moving the observations

Place sample names beside their points. Offset text when necessary while leaving points at their real coordinates. For colliding labels, move the text and use a restrained leader rather than dragging a circle into empty space. Leaders must clearly identify points without resembling a trend line connecting observations. If a label includes values, consistently write x before y and explain that order in the legend. Do not connect the five IDs in sequence, which would imply a path or time relationship. Hide labels to count points, then restore them and verify that each marker has exactly one traceable identity.

Explain overlaps without inventing positions

In a test copy, change E to (3,3), matching C, updating the list first. Two points now overlap. Do not move one arbitrarily simply to reveal five circles while leaving its coordinate label unchanged. A note can state that this location contains two samples; another explicitly explained count encoding is possible, but text is the basic method here. Jitter would need disclosure that it adjusts display positions rather than preserving exact coordinates, and is not assumed in this exercise. Restore the original data or save a separately named overlap version. Keep list and drawing aligned so a five-sample title never accompanies four unexplained visible points.

Describe distribution without claiming causation

You may observe that one point has the highest y or two have similar x values, but verify each observation against the list. Five invented samples establish no real-world law or causal claim merely because the overall pattern appears upward. Do not draw a smooth curve through the group and call it a fitted result; fitting requires a stated method and computation. Label the plot a fictional paired-value example whose source is your exercise list. For later real data, check sample scope, units and permission before deciding which interpretation is justified rather than mechanically transferring this explanation.

Audit the actual output from axes to points

Save the editable source and inspect the export: first axis ranges, units and ticks, then A–E coordinates, then the smallest labels and leaders. At actual display size, oversized markers can obscure positions; reduce their size without moving their centres. Retain full axis names and the example’s fictional status when using it in a report instead of cropping away the scale. Include data revision and dimensions in the filename and reopen the file to be shared, excluding leftover overlap-test notes or draft references. Future editing should use the editable source because a fixed-pixel preview cannot replace coordinate-object structure.

Return to the pair list when the plot reads incorrectly

If points appear to rise in sequence, check whether IDs were mistaken for x values. For a wrong height, check that each row’s pair remained together. An out-of-range point calls for checking the axis range, not dragging the value inside to hide it. Repair ambiguous text and leaders without moving observations. For circles missing after export, inspect fill, stroke and background contrast. After repairs, read both values for every point again and verify there are no unjustified connecting lines or fitting claims. A useful scatterplot traces every point to its original pair rather than merely resembling an expected trend.