Interpreting Slope and Intercept
A café wants to know how many customers it loses for every extra penny on a coffee; a tutor wants to
know how many marks an extra hour of revision really buys. Both answers are hiding inside two numbers
on a fitted line — and the value of a regression is only as good as your ability to say what they mean.
Fitting the regression
line gives you two numbers, a and
b. Getting them is the easy part. The real skill — the one that
separates a useful model from confident nonsense — is being able to say, in plain words, what those
two numbers mean in the real world.
In
\hat y = a + b x,
the slope b tells you how much y
changes per unit of x, and the intercept
a tells you y's predicted value when
x is zero. Read them right and the line speaks English; read
them wrong and a good model turns into a headline-grade blunder.
The slope b
The slope b is the change in the predicted
\hat y for a one-unit increase in x.
If a line predicts exam score from hours studied as
\hat y = 50 + 3x, the slope 3 says: each
extra hour of study is associated with 3 more points on the predicted score. A
negative slope means \hat y falls as x rises.
Notice the phrasing — "each extra hour is associated with." The slope reports how the
prediction moves; it does not, by itself, prove that studying causes the rise. But as a
description of the line, it is exact: step x up by one, and
\hat y steps up by b.
The intercept a
The intercept a is the predicted \hat y
when x = 0 — where the line crosses the vertical axis.
In \hat y = 50 + 3x it predicts a score of
50 for someone who studied zero hours.
But beware: x = 0 is often outside the range of the data,
or physically meaningless. A line fitted to adult heights and weights might report a "weight at
height zero" — a number with no sensible interpretation. Reading the intercept then is
extrapolation, and should be done with care or not at all.
Three worked readings
1) Reading a slope and intercept in English. A tutor fits
\hat y = 40 + 5\,x,
where x is hours studied and y is the exam
score (out of 100). The slope 5 reads:
"each extra hour of study predicts 5 more points on the exam." The
intercept 40 reads: "a student who studies
0 hours is predicted to score 40." Notice the units baked
into every sentence — points, hours — without them the numbers are mute.
2) Making a prediction. How does the same line score a student who studies
7 hours? Substitute:
\hat y = 40 + 5 \times 7 = 40 + 35 = 75. Predicted score:
75 out of 100.
3) A negative slope. A shop fits
\hat y = 500 - 8\,x,
where x is the price of a coffee (in pence above £1) and
y is cups sold per day. The slope -8 reads:
"each extra penny on the price predicts 8 fewer cups sold." The sign
carries the whole story here — a positive slope would mean raising the price sells more,
which would be a very strange café indeed.
Predict by plugging in
To predict, substitute a value of x into the line. Set the slope and
intercept, then slide the prediction point along the
x-axis: the marker rides the line and the readout shows
\hat y = a + b x at that x. Notice how a
steeper slope makes the prediction climb faster, and the intercept shifts the whole line up or
down. Push a negative slope and watch the prediction fall as
x grows.
- Slope b: the change in \hat y per one-unit increase in x.
- Intercept a: the predicted \hat y when x = 0 — meaningless if x = 0 is far outside the data.
- Predict by plugging an x into \hat y = a + b x.
The intercept is the honest answer to a question you may never actually want to ask: "what does the
line predict at x = 0?" Very often x = 0 is
far outside the data or downright impossible, and then the intercept is a mathematical artefact, not
a real-world fact.
Take a regression of a person's weight on their height, fitted from adults between
1.5 m and 2.0 m tall. The intercept is the predicted weight at "height = 0" —
a person no metres tall. It may even come out negative. That is not a discovery about
featherweight zero-metre people; it's the line being stretched to a place no data ever visited. The
slope is still perfectly meaningful ("each extra cm predicts so many more kg"), but the intercept
here is best treated as a bookkeeping number that positions the line — not as a prediction to
quote. Before you interpret an intercept literally, always ask: is
x = 0 a real, in-range value?
A slope written as a bare number is almost meaningless. "The slope is 5"
— five what, per what? "5 exam points per hour of study" is a
claim you can act on; "5" alone is just a number. The units are half the
meaning.
Units also explain why the same slope value can mean wildly different things. A slope of
5 "points per hour" and a slope of 5 "kg per
cm" and a slope of 5 "deaths per pound of budget cut" are all the number
5, and worlds apart in what they say. Change the units of x (measure
study in minutes instead of hours) and the slope's number changes even though the reality
hasn't — which is exactly why statisticians sometimes use standardised coefficients
(slopes rescaled so they can be compared across variables), or, more simply, why you should
always state the units.
Get the units — or the sign — wrong and the conclusion can flip. Real policy debates have
turned on someone reading a slope backwards, "proving" that a helpful programme was harmful because
a minus sign or a unit was mishandled. The arithmetic was fine; the interpretation was the
whole ballgame.
See it explained