Constant variance in AP Statistics

Constant variance (homoscedasticity) is the condition that the standard deviation of the y-values stays approximately the same across all values of x. On the AP Stats exam, you check it with a residual plot before doing a confidence interval or t-test for the slope of a regression line.

What is constant variance?

Constant variance, also called homoscedasticity, is the condition that the standard deviation of the response variable y does not change as x changes. In the CED's language, "the standard deviation for y, σy, does not vary with x." Picture the scatterplot as a cloud of points around the regression line. Constant variance means that cloud is roughly the same thickness everywhere, not skinny on the left and fat on the right.

You verify this condition by looking at the residual plot. If the residuals form a band of roughly even width across all x-values, the condition is met. If the residuals fan out into a funnel or megaphone shape (small spread at one end, big spread at the other), the condition is violated. The exam phrase for the violation is heteroscedasticity, but you'll mostly just describe what you see in the plot.

Why constant variance matters in AP® Statistics

Constant variance is one of the four conditions (linear, independent, normal, equal SD, the "LINE" mnemonic) that the CED requires you to verify before doing inference for slopes in Unit 9. It shows up explicitly in two learning objectives. AP Stats 9.2.B requires you to verify conditions before building a confidence interval for a slope, and AP Stats 9.4.C requires the same check before running a t-test for a slope. There's also a math reason behind the rule. The standard error formula for the slope, SE = s/(sx√(n-1)), uses a single value of s to estimate σ, the spread of y around the population regression line. That only makes sense if there IS a single spread. If the variance changes with x, one number can't honestly describe it, and your margin of error and p-value are built on a shaky estimate.

How constant variance connects across the course

Residual Plots (Unit 2)

You first met residual plots in Unit 2 as a tool for checking whether a linear model fits. Unit 9 reuses the exact same plot for a second job, checking that the spread of residuals stays even across x. One picture, two conditions.

Confidence Intervals for the Slope (Unit 9)

Constant variance is condition (b) in the CED's checklist for building the interval b ± t*(SEb). If the funnel shape appears, the standard error in your margin of error is unreliable, so the whole interval is suspect.

t-Test for a Slope (Unit 9)

The significance test in Topic 9.4 requires the same condition. Before testing H₀: β = β₀, you check the residual plot for even spread, because the test statistic divides by a standard error that assumes one constant σ.

Independence Condition (Units 6-9)

Constant variance and independence are both items on the slope-inference checklist, but they're checked differently. Independence comes from the design (random sample plus the 10% condition), while constant variance comes from looking at the residual plot.

Is constant variance on the AP® Statistics exam?

This term lives in the "check conditions" step of slope inference. Multiple-choice questions love showing or describing a residual plot with a funnel shape (residuals spreading out as x increases) and asking which condition is violated. The answer is constant variance, even though the plot might tempt you toward linearity. Other MCQs flip it, describing a residual plot with random scatter and asking which conditions ARE satisfied, or asking which condition is NOT required for a slope interval. On an FRQ, expect to state and verify all conditions before computing a confidence interval or running a t-test for a slope. For constant variance, that means writing something like "the residual plot shows roughly equal spread across all values of x, so the equal-SD condition is met." Just naming the condition without pointing to the plot won't earn full credit.

Constant variance vs Linearity condition

Both conditions are checked with the same residual plot, which is exactly why they get mixed up. A curved or U-shaped pattern in the residuals violates linearity, meaning a line is the wrong model. A funnel or megaphone shape violates constant variance, meaning the spread of y changes with x. The shape of the pattern tells you which condition failed. Bend means linearity, fan means constant variance.

Key things to remember about constant variance

• Constant variance (homoscedasticity) means the standard deviation of y is approximately the same for every value of x.

• You check it with a residual plot, and a roughly even band of scatter means the condition is met.

• A funnel or fan shape in the residual plot, with residuals spreading out as x increases, is the classic sign the condition is violated.

• It is a required condition for both confidence intervals for a slope (Topic 9.2) and t-tests for a slope (Topic 9.4).

• The condition matters because the slope's standard error uses one value of s to estimate σ, which only works if the spread of y is actually constant.

• Don't confuse it with linearity. A curved residual pattern violates linearity, while a fanning pattern violates constant variance.

Frequently asked questions about constant variance

What is constant variance in AP Stats?

It's the condition that the standard deviation of the y-values stays approximately the same across all values of x, also called homoscedasticity. It's one of the conditions you must verify before doing a confidence interval or t-test for the slope of a regression line in Unit 9.

How do you check the constant variance condition?

Look at the residual plot. If the residuals form a band of roughly equal width across all x-values, the condition is satisfied. If they fan out into a funnel shape, it's violated.

Does a funnel-shaped residual plot violate linearity?

No, it violates constant variance, not linearity. A linearity violation shows up as a curved or U-shaped pattern, while a funnel shape (residuals spreading out as x increases) means the spread of y is changing with x.

Is constant variance the same as homoscedasticity?

Yes, they're two names for the same condition. "Homoscedasticity" is the formal statistics term, and its violation is called heteroscedasticity, but the AP exam usually describes it as equal or constant standard deviation of y across x.

Why does slope inference need constant variance?

The standard error of the slope, SE = s/(sx√(n-1)), uses a single value of s to estimate σ, the spread of y around the population regression line. If the spread changes with x, no single number describes it, so your margin of error and p-value become unreliable.