Check Error Bars and Linear Trendlines in Numbers Charts

Quick answer and scope

A trendline summarizes a pattern; error bars display a chosen uncertainty convention. They answer different questions. This exercise keeps both visible without treating either as proof that measurements are reliable.

The Apple Mac guide identifies Numbers 15.4. These are documentation-qualified Mac instructions, checked October 22, 2026, not iPad, iPhone or iCloud procedures. No native Numbers execution, screenshot or save/reopen test was available. The arithmetic below was independently calculated, not read from a rendered Numbers chart.

Build a deliberately simple scatter example

In a new practice spreadsheet, create a table with one header row and no footer. Enter:

Cell	Value	Cell	Value
A1	Input	B1	Response
A2	1	B2	4
A3	2	B3	6
A4	3	B4	8

Use the scatter-chart setup with A as the numeric x values and B as the single y series. Verify the three plotted pairs: (1,4), (2,6), (3,8). Do not use a category line chart as a substitute for this exercise.

The data are invented teaching values. The following uncertainty amount is an assigned visual exercise, not an estimate computed from repeated observations.

Add and inspect the markings

Select the chart and open Format → Series. Expand Error Bars. For this scatter chart, use the Y Axis Error Bars controls: choose bars on both sides, then fixed values, and set the amount to 1. Leave x error bars off. Apple documents separate x/y controls and fixed-value or percentage calculations. Chart markings reference.

The mathematical y endpoints should be 3–5, 5–7 and 7–9. Each interval is two units wide, centered on its original response. A percentage setting would mean something different: 10% of 4 is 0.4, not 1. Check the selected calculation before interpreting the display.

For an optional fit check, expand Trendlines in Series. Apple documents a type selector plus Show Equation and Show R² Value, but the retrieved guide does not enumerate the available type labels. Continue with the calculation comparison below only if your app actually offers a linear/straight-line model; otherwise stop this optional portion. Do not substitute another model. Documented controls.

Independently check the fit

These are independently calculated linear-model expectations, not an observed Numbers equation or verified menu selection. The fixed-error-bar exercise above remains useful without this optional comparison.

For these three points, every response equals 2×Input+2. Therefore the expected straight-line slope is 2, intercept 2 and R² is 1. Display rounding or different equation typography does not change that relationship. For worksheet calculations, see SLOPE, INTERCEPT and RSQ.

Change only B3 from 6 to 7. The points are now (1,4), (2,7), (3,8). Independently calculated least-squares expectations are:

• Slope: 2.
• Intercept: 7/3, approximately 2.333333.
• R²: 12/13, approximately 0.923077.
• Fixed y intervals: 3–5, 6–8 and 7–9.

This is a useful diagnostic: the middle point and its error bar move, but the fixed error amount remains 1. The fitted slope happens to stay unchanged; the intercept and R² do not. Restore B3 to 6 and verify the original three points and fit again. Save a working copy only after checking the intended state.

Troubleshooting and interpretation

If some bars are hidden, Apple’s Auto-Fit option may be suppressing overlap; deselect it to inspect all bars. Settings apply to the selected series, so inspect that selection before changing another. Remove error bars by selecting a bar and choosing None from the Error Bars menu. These controls are documented for supported chart types, not every chart. Apple guidance.

If the numeric checks disagree, first verify the three source rows, x/y assignment and chosen linear model. Do not adjust source values merely to make a displayed equation look right. R²=1 in this constructed example is not evidence of causation, useful forecasting or correctly estimated uncertainty. The bars were assigned, not calculated as standard deviations or confidence intervals.