The Same Data Can Look Different
In “Reading Values From a Scatterplot,” you learned to read each coordinate from its own axis scale. Those scales do more than help us estimate values: they also shape how the overall pattern looks. A tall, narrow plot can make a pattern seem steeper than a wide, short one, and a restricted axis range can make differences appear larger. The paired observations themselves have not changed.
This matters when you describe strength. As in “Judging Strength of an Association,” strength concerns how closely the points follow the overall pattern. But our visual judgment can be influenced by the graph’s layout. A stretched or compressed display may make the scatter around that pattern seem different, even though every point has the same coordinates.
An axis scale tells you how data values map to distances on the page or screen. The displayed range is the interval shown on an axis, from its lower limit to its upper limit. If the same graph height represents a larger range of response values, each unit gets less vertical space and the plotted pattern is compressed vertically. If the displayed range is smaller, each unit gets more space and the pattern is stretched vertically.
The graph’s physical proportions matter too. Even with identical axis labels and ranges, making the plot taller or wider changes the visual angle and shape of the pattern. That can affect how steep or compact an association seems. These are reasons to inspect the scales and graph dimensions before relying on an impression.
A Scale Audit Before Judging Strength
A quick scale audit helps separate what the data show from what the display emphasizes. First, check the variables, units, and numerical limits on both axes. Then compare the span occupied by the observations with the full displayed range. Finally, notice whether the graph is unusually tall, wide, or tightly cropped.
Note the lower and upper limits, the intervals between tick marks, and the units. Do not assume equal-looking distances represent equal amounts on the two axes.
Compare the smallest and largest observed values with the displayed range. A larger displayed range leaves more empty space and compresses the observations.
A tall or wide plotting area changes the apparent geometry. A restricted range can magnify differences, so check that all relevant observations are included.
Use the direction, form, unusual features, and strength in context, while avoiding claims that a visual change means the data changed.
One useful diagnostic is the fraction of the displayed axis range occupied by the observed values. It is not a measure of association or strength. It simply helps you see whether the graph gives the data a lot of room or compresses them into a smaller portion of the display.
For example, if observations range from 18 to 34 and the vertical axis runs from 0 to 36, they use \(16/36\), or about 44.4%, of that axis. If the axis instead runs from 0 to 72, the same observations use \(16/72\), or about 22.2%. The data are identical, but they occupy only half as much of the displayed vertical range.
One Pattern, Two Vertical Scales
Worked Example: Weekly Training and Active Minutes
Imagine a fictional scatterplot relating weekly training hours, \(x\), to active minutes during a workout, \(y\), for five participants. The invented data are:
| Training hours, \(x\) | Active minutes, \(y\) |
|---|---|
| 1 | 18 |
| 2 | 25 |
| 3 | 23 |
| 4 | 34 |
| 5 | 31 |
Compare two displays with the same width of 600 pixels and height of 360 pixels. In both, the horizontal axis runs from 0 to 6 hours. In Display A, the vertical axis runs from 0 to 36 minutes; in Display B, it runs from 0 to 72 minutes.
Check that the data fit both displays. The smallest active-minute value is 18 and the largest is 34. Both values, and every value between them, lie within 0 to 36 and within 0 to 72. No plotted observation is cut off in either display.
Compare the vertical scales. Display A has 360 pixels for 36 minutes, or 10 pixels per minute. Display B has the same height for 72 minutes, or 5 pixels per minute. The observed vertical range is \(34-18=16\) minutes. It occupies \(16/36\), about 44.4%, of Display A’s vertical range, and \(16/72\), about 22.2%, of Display B’s.
Explain the visual effect. The observed rise from 18 to 31 minutes between the first and last training-hour values is 13 minutes. On Display A, that corresponds to \(13\times10=130\) pixels vertically; on Display B, it corresponds to \(13\times5=65\) pixels. With the same horizontal scale, the pattern appears steeper in Display A and flatter in Display B.
Conclude carefully. Both displays show the same five pairs and the same general positive direction, with some variation around the pattern. A viewer might get a different impression of its visual steepness or compactness, but the observations and their association have not changed. Neither display alone establishes that the association is stronger.
This comparison shows why “the pattern looks steeper” and “the association is stronger” are not interchangeable statements. Steepness concerns how much the response tends to change for a given change in the explanatory variable. Strength concerns how closely points follow the overall pattern. A display can change the apparent angle without changing the points.
When a Restricted Axis Magnifies Differences
Scatterplots do not always need to begin at zero. A carefully chosen range can make the region containing the observations easier to inspect. But a restricted range also gives each unit more visual space. If readers do not notice the limits, small differences can look dramatic. A narrow range is not automatically misleading; the risk comes from failing to notice or communicate what it shows.
Worked Example: Zooming In on the Training Data
Use the same five training-hour and active-minute pairs from the previous example. Consider a 400-pixel-high display with the vertical axis running from 0 to 36 minutes, and then a second 400-pixel-high display running from 16 to 36 minutes.
Check the observations against the restricted range. The observed active-minute values are 18, 25, 23, 34, and 31. The minimum is 18 and the maximum is 34, so every point remains within the restricted range of 16 to 36. The restriction does not remove any of these five observations.
Calculate the scale per minute. With a 0-to-36 range, 400 pixels represent 36 minutes, or about \(400/36=11.1\) pixels per minute. With a 16-to-36 range, 400 pixels represent 20 minutes, or \(400/20=20\) pixels per minute.
Compare the displayed spread. The observations span \(34-18=16\) minutes. In the first display they use \(16/36\), about 44.4%, of the axis height, which is about \(400(16/36)=177.8\) pixels. In the restricted display they use \(16/20=80\%\), or \(400(16/20)=320\) pixels.
Interpret the effect. The differences in active minutes occupy much more of the second graph, so the points’ vertical spread and changes in \(y\) look more prominent. The points still have the same coordinates. The zoomed display may be useful for examining detail, but a reader should see the axis limits and understand that the vertical differences have been magnified.
A restricted range can emphasize variation in \(y\); it does not automatically make the association stronger or weaker. Strength still depends on how the points follow the overall pattern. If the display makes that judgment difficult, describe the plotted pattern cautiously and refer to the data or a suitable numerical summary rather than treating visual drama as proof.
Changing the Horizontal Scale Also Matters
The same principle applies to the explanatory-variable axis. Expanding its displayed range compresses the observations horizontally; narrowing the range expands them horizontally. This can change the apparent slope and the cloud’s shape. For a fair visual comparison, use consistent axis limits and plot proportions when the goal is to compare patterns across groups or settings.
Worked Example: Expanding the Training-Hour Axis
Again use the same five invented data pairs. Keep the vertical axis at 0 to 36 minutes and use a plot 600 pixels wide. Compare a horizontal axis from 0 to 6 hours with one from 0 to 10 hours.
Check that both horizontal ranges include the observations. The training-hour values run from 1 to 5, so all five values lie within both 0 to 6 and 0 to 10 hours. No point is excluded by either horizontal scale.
Compare the horizontal scale per hour. With a width of 600 pixels, the 0-to-6 axis provides \(600/6=100\) pixels per hour. The 0-to-10 axis provides \(600/10=60\) pixels per hour.
Calculate the width used by the observations. The observed range is \(5-1=4\) hours. On the first display, that uses \(4/6\) of the width, or \(600(4/6)=400\) pixels. On the second, it uses \(4/10\) of the width, or \(600(4/10)=240\) pixels.
Interpret the effect. The data appear more spread out horizontally on the 0-to-6 display and more compressed on the 0-to-10 display. Since the vertical scale did not change, the same rise in active minutes can look steeper relative to the horizontal distance in the compressed display. This is an appearance change, not a change in training hours or active minutes.
When comparing two scatterplots, do not conclude that one association is stronger just because its points fill more of the plotting area. Check that the axes use the same ranges and units and that the plot proportions are comparable. If they are not, describe the limitation before comparing the visual patterns.
What Axis Changes Do—and Do Not—Change
Changing axis limits or graph proportions changes the way points are drawn on the page. It does not change which observations have larger or smaller \(x\)- or \(y\)-values, the coordinates of any point, or the actual data set. A different scale cannot turn a positive association into a negative one in the data. It can, however, make the pattern’s angle or visible spread seem different.
Also distinguish axis scale from changing the values themselves. Relabeling an axis without changing the data is a display choice. Converting every measurement to another unit changes the numerical labels but preserves the underlying observations and their ordering. Either way, read the axis labels and units before comparing distances or slopes.
If a later numerical summary is calculated from the same paired values, merely redrawing the graph does not alter that summary. The display is not new evidence about the relationship; it is a different visual representation of the same evidence. This is why an explanation of strength should focus on how closely points follow their pattern, not just on the angle or amount of page space they occupy.
Common Mistakes and AP Exam Tips
- Calling a steeper-looking pattern stronger: Steepness and strength describe different features. A full-credit response identifies the visual change, then explains that the same points remain and the apparent angle alone does not establish stronger association.
- Ignoring the axis limits: A graph that starts at 16 does not show the same vertical range as one that starts at 0. Name or check the limits before comparing how much space the observations occupy.
- Assuming a restricted axis is always wrong: A restricted range may help show detail if it includes the relevant values and is clearly labeled. Explain its magnifying effect rather than automatically rejecting the graph.
- Claiming data were omitted without checking: Compare every observed minimum and maximum with the displayed limits. In the examples, the active-minute values from 18 through 34 fit within both 0-to-36 and 16-to-36 ranges.
- Comparing plots with different scales as if they were identical: Check units, axis ranges, and physical proportions. If they differ, say that the visual comparison is limited.
- Confusing the graph’s appearance with a change in association: The plotted coordinates do not change when the image is stretched. State explicitly that the display changes, not the observations.
A strong AP response is specific: “The second graph uses a larger vertical range, so the same active-minute values occupy less of its height and the pattern appears flatter. The five paired observations are unchanged, so the visual difference does not mean the association itself changed.” That answer names the scale choice, its visual effect, and the important limitation.
Check Your Understanding
Use the scale ideas in this tutorial to answer each question.
- Five response values range from 12 to 24. What fraction of a vertical axis from 0 to 30 do they occupy? What fraction of an axis from 0 to 60?
- A scatterplot is made taller while both axis limits and all points remain fixed. Name one visual feature that may change and one thing that does not change.
- A scatterplot’s vertical axis runs from 20 to 40. The observed response values range from 18 to 37. What should you check before interpreting the graph?
- Two scatterplots show similar patterns, but one uses a horizontal range of 0 to 5 and the other 0 to 10. What scale information should you check before comparing their apparent strength?
- Why is “the points look more spread out, so the association must be stronger” an incomplete conclusion when two graphs use different axis ranges?