The Mean Point on a Regression Line
In “Computing Slope From \(r\) and Summary Statistics,” you used the sample means and slope to calculate the intercept. That calculation works because a least-squares regression line has an important property: it passes through the point whose coordinates are the sample means of the predictor and response, \((\bar{x},\bar{y})\).
This point summarizes the center of the observed \(x\)-values and the center of the observed \(y\)-values. The property gives us a direct way to find the intercept when the slope and sample means are known. It also gives us a quick check on an equation: when \(x=\bar{x}\), the line must predict \(\hat{y}=\bar{y}\).
The mean point is a point on the fitted line, but it does not have to represent an actual individual in the sample. For example, \(\bar{x}\) might be a value that no one in the sample has. The statement is about the line and the sample means, not about a particular observed case.
Why the Line Passes Through the Means
Recall the sample regression equation \(\hat{y}=a+bx\), where \(x\) is the predictor, \(\hat{y}\) is the predicted response, \(a\) is the intercept, and \(b\) is the slope. The least-squares line is chosen to minimize the sum of the squared residuals. For a least-squares line with an intercept, the residuals sum to zero.
A residual is an observed response minus its predicted response: \(y-\hat{y}\). If the residuals sum to zero, then their average is zero, too. So, on average, the observed \(y\)-values equal the predicted \(y\)-values. The average observed response is \(\bar{y}\); the average predicted response from the line is \(a+b\bar{x}\). Equating these averages shows why the line goes through the mean point.
The same result gives the intercept formula. Rearrange \(\bar{y}=a+b\bar{x}\) to isolate \(a\). This is the relationship used in the previous tutorial, now understood as a consequence of the regression line passing through the means.
You can also see the property by substituting the intercept formula into the line at \(x=\bar{x}\):
This substitution is a useful calculation check. It does not replace the need to use the correct predictor and response means, but it can expose an arithmetic or sign error in the intercept.
A Reliable Way to Find or Check the Intercept
If a problem gives a slope and the two sample means, use the mean point to find the intercept. Write the formula before inserting numbers, and keep a negative slope in parentheses. Once you have the equation, substitute \(\bar{x}\) into it and check that the result is \(\bar{y}\).
Confirm which variable is the predictor \(x\), which is the response \(y\), and which means are \(\bar{x}\) and \(\bar{y}\).
Substitute the known slope and means into \(a=\bar{y}-b\bar{x}\) to calculate the intercept.
Write \(\hat{y}=a+bx\), then verify that \(a+b\bar{x}=\bar{y}\).
The slope \(b\) has response units per predictor unit; the intercept \(a\) has response units. In the formula \(b\bar{x}\), the predictor units cancel, leaving response units, so it can be subtracted from \(\bar{y}\). This units check also helps you avoid mixing up the two means.
Worked Calculations
Worked Example: River Flow and Dissolved Oxygen
A fictional environmental monitoring team records \(x\), river flow in cubic meters per second, and \(y\), dissolved oxygen in milligrams per liter. For the sample, \(\bar{x}=12\), \(\bar{y}=8.4\), and the least-squares slope is \(b=-0.15\) milligrams per liter per cubic meter per second. Find the intercept and write the regression equation.
The regression line passes through \((12,8.4)\), so put the means and slope into the intercept formula:
The intercept is \(10.2\) milligrams per liter. Thus, the equation is \(\hat{y}=10.2-0.15x\), with \(x\) measured in cubic meters per second and predicted \(y\) measured in milligrams per liter. The line predicts dissolved oxygen to be \(0.15\) milligrams per liter lower for each one-unit increase in river flow. The intercept describes the line’s prediction at zero river flow; whether that is useful depends on the context and the range of observed flows.
Check the mean point by substituting \(x=12\):
The prediction equals \(\bar{y}=8.4\), as it must. Notice that the negative slope made the intercept larger than the mean response: the formula subtracts the negative product \(b\bar{x}\).
Worked Example: Elevation and Afternoon Temperature
A fictional mountain survey uses \(x\), elevation in hundreds of meters above a reference level, to predict \(y\), afternoon temperature in degrees Celsius. The sample summaries are \(\bar{x}=4\) hundreds of meters and \(\bar{y}=18\) degrees Celsius. The fitted slope is \(b=-1.6\) degrees Celsius per hundred meters. Find the intercept and check the fitted line at the sample mean elevation.
Because the regression line passes through \((4,18)\), calculate:
The intercept is \(24.4\) degrees Celsius, and the regression equation is \(\hat{y}=24.4-1.6x\). Its slope says that for each additional hundred meters of elevation, the line predicts an afternoon temperature \(1.6\) degrees Celsius lower. The intercept is the predicted temperature at the reference elevation \(x=0\); its practical usefulness depends on whether that elevation is relevant to the survey.
Substitute the mean elevation into the equation:
This matches the sample mean temperature. The check confirms that the equation goes through \((4,18)\). It does not mean that every observation at the mean elevation has a temperature of exactly \(18\) degrees Celsius; individual responses can differ from the line’s prediction.
Worked Example: Charging Time and Battery Capacity
A fictional consumer-electronics sample compares \(x\), battery capacity in thousands of milliamp-hours, with \(y\), charging time in hours. The sample means are \(\bar{x}=5\) and \(\bar{y}=2.8\), and the fitted slope is \(b=0.30\) hours per thousand milliamp-hours. Find the intercept and use the mean point to check the equation.
Use the predictor mean, response mean, and slope in that order:
The fitted equation is \(\hat{y}=1.3+0.30x\). The intercept has units of hours, and the slope has units of hours per thousand milliamp-hours. In context, the equation predicts charging time to be \(0.30\) hours greater for each additional thousand milliamp-hours of battery capacity. At \(x=0\), it predicts \(1.3\) hours, though that prediction may not be practically meaningful if zero capacity is not represented or is impossible for the devices considered.
Check the sample mean point:
The result is the sample mean charging time, so the equation passes through \((5,2.8)\). The mean point does not assert that a device with exactly mean capacity must have exactly mean charging time; it is a property of the fitted line.
Common Mistakes and AP Exam Tips
On an AP response, show the relationship that justifies the calculation, not only the final number. A clear answer states that the least-squares line passes through \((\bar{x},\bar{y})\), uses \(a=\bar{y}-b\bar{x}\), and then reports the equation with the predictor and response identified.
- Reversing the means. Use \(a=\bar{y}-b\bar{x}\), not \(\bar{x}-b\bar{y}\). The intercept is measured in response units, so its calculation must begin with the response mean.
- Dropping parentheses around a negative slope. If \(b\) is negative, write \(a=\bar{y}-(-|b|)\bar{x}\). Subtracting a negative product adds it.
- Claiming the mean point is an observed case. The line passes through the point of the sample means, but that point need not correspond to any one individual in the data.
- Confusing a prediction with an actual response. At \(x=\bar{x}\), the line predicts \(\bar{y}\). This does not say that every observation has response \(\bar{y}\) or that a particular case must lie on the line.
- Giving an intercept without units or context. The intercept has response units and is the line’s predicted response when \(x=0\). As in “When the Intercept Has No Practical Meaning,” explain cautiously if zero is impossible, irrelevant, or outside the data range.
- Rounding too early. If the slope or means have more precise values available, use them in the calculation and round the reported intercept and equation at the end.
A concise full-credit calculation might say: “The least-squares line passes through \((\bar{x},\bar{y})\), so \(a=\bar{y}-b\bar{x}=[\text{substitution}]=[\text{value}]\). Therefore, the fitted line is \(\hat{y}=a+bx\). Substituting \(x=\bar{x}\) gives \(\hat{y}=\bar{y}\), which checks the result.” Add an interpretation of the intercept when the question asks for one.
Check Your Understanding
Use the mean-point property in each question, and keep the predictor and response roles clear.
- A fitted line has slope \(b=2.5\), \(\bar{x}=4\), and \(\bar{y}=16\). Find the intercept and verify that the line passes through the sample mean point.
- A sample has \(\bar{x}=8\), \(\bar{y}=11\), and \(b=-0.75\). Calculate the intercept. Show how the negative slope affects the arithmetic.
- In words, explain why the least-squares regression line passes through \((\bar{x},\bar{y})\), using the fact that the residuals sum to zero.
- Does the mean point on a regression line have to be an actual observation? Explain.
- A student calculates the intercept, then substitutes \(\bar{x}\) into the equation and obtains a value different from \(\bar{y}\). Name two possible calculation errors the student should check.