Slope vs y-intercept in regression

By Jude Wallis · Published

In the regression line y^=a+bx\hat{y} = a + bx, the slope bb is the predicted change in the response for each one-unit increase in the explanatory variable. The y-intercept aa is the predicted response when the explanatory variable equals 0. The slope is a rate, and the intercept is a starting value.

AP Statistics: Unit 5 (topics 5.3 Linear Regression Models, 5.5 Least-Squares Regression). The regression equation and prediction are in topic 5.3, and interpreting the slope and the y-intercept of the least-squares line is in topic 5.5, both in Unit 5 of the Fall 2026 AP Statistics course. The formula sheet lists only the form y-hat = a + bx, and the coefficients are normally read from technology.

Slope vs y-intercept: the short answer

A least-squares regression line is written y^=a+bx\hat{y} = a + bx, where y^\hat{y} ("y-hat") is the predicted value of the response variable and xx is the explanatory variable. The two numbers in the equation do different jobs.

The slope bb tells you how the prediction changes as xx changes. It is a rate, with units of yy per unit of xx. The y-intercept aa tells you where the line sits when x=0x = 0. It is a single predicted value, in the units of yy.

The slope almost always has a useful meaning in context. The intercept has one only when x=0x = 0 is a realistic value near the data. The AP formula sheet writes the equation with the intercept first, so in y^=61.4+5.3x\hat{y} = 61.4 + 5.3x the slope is 5.3, the number multiplying xx.

What the slope tells you

The slope is the predicted change in the response for a one-unit increase in the explanatory variable. A sentence frame that works for any context: for each additional one unit of the explanatory variable, the predicted response increases (or decreases) by the slope, stated with its units.

If y^=61.4+5.3x\hat{y} = 61.4 + 5.3x predicts a test score from hours of tutoring, the slope says that each additional hour of tutoring is associated with a predicted score 5.3 points higher.

Three facts about the slope:

  • Its sign gives the direction of the association, and it always matches the sign of the correlation rr.
  • Its size depends on the units. Measure tutoring in minutes instead of hours and the slope becomes 5.360≈0.088\frac{5.3}{60} \approx 0.088 points per minute, with no change in the relationship.
  • It describes the line's prediction, so the word "predicted" belongs in the interpretation. It does not promise that any individual gains exactly that amount.

The full wording is covered in how to interpret the slope of a regression line.

What the y-intercept tells you

The y-intercept is the predicted response when x=0x = 0. The sentence frame: when the explanatory variable is 0, the predicted response is the intercept, stated with its units.

For y^=61.4+5.3x\hat{y} = 61.4 + 5.3x, a student with 0 hours of tutoring has a predicted score of 61.4 points. That reading makes sense if the data include students with little or no tutoring.

Often it does not make sense. If a line predicts adult weight from height, x=0x = 0 means a person who is 0 inches tall. No such person exists, the value is far outside the data, and the intercept may come out negative. Using the line there is extrapolation. The intercept is still needed to place the line at the correct height over the data. It simply has no interpretation in context.

The intercept is tied to the slope by the fact that the least-squares line passes through the point of means (xˉ,yˉ)(\bar{x}, \bar{y}):

a=yˉ−bxˉa = \bar{y} - b\bar{x}

The details are in how to interpret the y-intercept of a regression line.

The differences side by side

FeatureSlope bbY-intercept aa
Position in y^=a+bx\hat{y} = a + bxMultiplies xxStands alone
What it isPredicted change in yy per one-unit increase in xxPredicted yy when x=0x = 0
UnitsUnits of yy per unit of xxUnits of yy
On the graphSteepness and direction of the lineWhere the line crosses the vertical axis at x=0x = 0
SignSame as the sign of rrNo link to the sign of rr
Meaningful in contextNearly alwaysOnly when x=0x = 0 is realistic and near the data
From summary statisticsb=r⋅sysxb = r \cdot \frac{s_y}{s_x}a=yˉ−bxˉa = \bar{y} - b\bar{x}

The formulas in the last row show the order of work: the slope is found first, and the intercept is then whatever value makes the line pass through (xˉ,yˉ)(\bar{x}, \bar{y}).

When each one matters

Use the slope to answer questions about change: how much more, how much less, per unit. Any comparison between two xx values uses the slope alone. Two students who differ by 2 hours of tutoring have predicted scores that differ by 2×5.3=10.62 \times 5.3 = 10.6 points, and the intercept cancels out of that subtraction.

Use the intercept, together with the slope, to predict a single value. A prediction at x=3x = 3 needs both: 61.4+5.3(3)=77.361.4 + 5.3(3) = 77.3.

Interpret the intercept in words only when asked and only when x=0x = 0 makes sense. When it does not, say so: state that the intercept is the predicted response at x=0x = 0, and that this value is outside the range of the data or impossible in context.

The classic mix-up and how to avoid it

The most common error is swapping the two numbers when reading an equation or computer output. School algebra writes y=mx+by = mx + b, with the slope first and bb as the intercept. Statistics writes y^=a+bx\hat{y} = a + bx, with the intercept first and bb as the slope. The letter bb changes jobs. Ignore the order and the letters, and find the number attached to xx. That number is the slope. In computer output, the slope is in the row labeled with the explanatory variable's name, and the intercept is in the row labeled Constant or Intercept.

The second error is interpreting the slope as a value of yy, as in "the predicted score is 5.3". The slope is a change in yy for a change in xx, so its interpretation always mentions both variables.

The third error is forcing a story onto an intercept that has none. A predicted weight of -97.5 pounds at a height of 0 inches should be reported as having no meaningful interpretation.

Finding and interpreting both from data

Five students received 0, 1, 2, 3, and 4 hours of tutoring and scored 62, 66, 73, 75, and 84 points on a test. Find the least-squares line and interpret the slope and the y-intercept in context.

  1. Find the means: xˉ=0+1+2+3+45=2\bar{x} = \frac{0+1+2+3+4}{5} = 2 hours and yˉ=62+66+73+75+845=3605=72\bar{y} = \frac{62+66+73+75+84}{5} = \frac{360}{5} = 72 points.

  2. Find the deviations from the means. For xx: -2, -1, 0, 1, 2. For yy: -10, -6, 1, 3, 12.

  3. Multiply the paired deviations and add: (−2)(−10)+(−1)(−6)+(0)(1)+(1)(3)+(2)(12)=20+6+0+3+24=53(-2)(-10) + (-1)(-6) + (0)(1) + (1)(3) + (2)(12) = 20 + 6 + 0 + 3 + 24 = 53.

  4. Square the xx deviations and add: 4+1+0+1+4=104 + 1 + 0 + 1 + 4 = 10.

  5. Compute the slope: b=5310=5.3b = \frac{53}{10} = 5.3 points per hour.

  6. Compute the intercept from the point of means: a=yˉ−bxˉ=72−5.3(2)=61.4a = \bar{y} - b\bar{x} = 72 - 5.3(2) = 61.4 points.

  7. Write the line: y^=61.4+5.3x\hat{y} = 61.4 + 5.3x. Interpret the slope: for each additional hour of tutoring, the predicted test score increases by 5.3 points. Interpret the intercept: a student with 0 hours of tutoring has a predicted score of 61.4 points. Here x=0x = 0 is one of the observed values, so the intercept is meaningful.

The line is y^=61.4+5.3x\hat{y} = 61.4 + 5.3x. The slope of 5.3 means the predicted score rises 5.3 points for each additional hour of tutoring. The y-intercept of 61.4 is the predicted score for a student with no tutoring, a sensible value because 0 hours is within the data.

An intercept with no meaning in context

For a group of adults, height has mean 66 inches and standard deviation 4 inches, weight has mean 150 pounds and standard deviation 25 pounds, and the correlation between height and weight is r=0.6r = 0.6. Find the least-squares line for predicting weight from height, interpret the slope, and explain what the intercept does and does not tell you.

  1. Compute the slope from the summary statistics: b=r⋅sysx=0.6×254=3.75b = r \cdot \frac{s_y}{s_x} = 0.6 \times \frac{25}{4} = 3.75 pounds per inch.

  2. Compute the intercept: a=yˉ−bxˉ=150−3.75(66)=150−247.5=−97.5a = \bar{y} - b\bar{x} = 150 - 3.75(66) = 150 - 247.5 = -97.5 pounds.

  3. Write the line: y^=−97.5+3.75x\hat{y} = -97.5 + 3.75x.

  4. Interpret the slope: for each additional inch of height, the predicted weight increases by 3.75 pounds.

  5. Read the intercept literally: a person 0 inches tall has a predicted weight of -97.5 pounds. A height of 0 is impossible and far below every height in the data, and a negative weight is impossible, so the intercept has no meaningful interpretation here.

  6. Show that the intercept still does its job. At x=70x = 70 inches the prediction is −97.5+3.75(70)=−97.5+262.5=165-97.5 + 3.75(70) = -97.5 + 262.5 = 165 pounds, a reasonable value. At the mean height, −97.5+3.75(66)=150-97.5 + 3.75(66) = 150 pounds, the mean weight.

The line is y^=−97.5+3.75x\hat{y} = -97.5 + 3.75x. The slope means predicted weight rises 3.75 pounds per additional inch of height. The intercept of -97.5 pounds has no meaningful interpretation, because a height of 0 inches is impossible and far outside the data, but the line needs it to give sensible predictions such as 165 pounds at 70 inches.

Frequently asked questions

Which number is the slope in y-hat = a + bx?

The slope is bb, the number multiplied by xx. The intercept is aa, the number standing alone. This is the reverse of the letter order in the algebra form y=mx+by = mx + b, so identify the slope by finding the coefficient of xx.

Can the y-intercept be negative when the slope is positive?

Yes. The sign of the intercept is unrelated to the sign of the slope. The intercept equals the mean of yy minus the slope times the mean of xx, so a positive slope with a large mean of xx often produces a negative intercept.

Do I always have to interpret the y-intercept?

Interpret it when a question asks, and say whether the interpretation makes sense. If x=0x = 0 is impossible or far outside the data, state that the intercept is the predicted response at x=0x = 0 and that it has no meaningful interpretation in this context.