Linear Regression Models
Introduced| English |
|---|
| regression line/rɪˈɡreʃn laɪn/ |
| extrapolation/ekˈstræpəleɪʃn/ |
A line through the cloud
- When a scatterplot looks linear, summarize it with a regression line 回归线.
- We write the predicted response as $\hat{y} = a + bx$ (the hat means "predicted").
- $a$ is the $y$-intercept, $b$ is the slope — the line's recipe.
- The line lets us predict $y$ from any $x$ and describe the trend precisely.
Reading the slope
- The slope $b$ is the predicted change in $y$ for each $1$-unit increase in $x$.
- In context: "each extra hour of study predicts about $b$ more points."
- Its sign matches the direction of the association (positive or negative).
- The slope is where almost all the interpretation credit lives on the exam.
Reading the intercept
- The intercept $a$ is the predicted $y$ when $x = 0$.
- Sometimes meaningful ("baseline score with zero study"), sometimes not.
- If $x = 0$ is far outside the data, the intercept is just a mathematical anchor.
- Interpret it in context — and be honest when $x=0$ makes no real sense.
Don't extrapolate
- Use the line only within the range of the observed $x$ values.
- Extrapolation 外推 — predicting far beyond the data — is unreliable.
- The linear pattern may simply not continue out there.
- A model that fits students studying $0$–$8$ hours says nothing trustworthy about $40$ hours.
Interpret the slope as a predicted change, not an actual one: "each extra hour is associated with about $b$ more points," not "causes." And never trust a prediction from extrapolation — plugging in an $x$ far outside the data range gives a number the data can't support, even if the arithmetic works.
For study hours, $\hat{y} = 55 + 4x$ (score predicted from hours).
- Slope $4$: each extra hour of study predicts about $4$ more points.
- Intercept $55$: a student who studies $0$ hours is predicted to score $55$.
- Predict at $x=3$: $\hat{y} = 55 + 4(3) = 67$ points — safely inside the data range.
A least-squares regression line $\hat{y} = a + bx$ models a linear trend. The slope $b$ is the predicted change in $y$ per $1$-unit increase in $x$; the intercept $a$ is the predicted $y$ at $x=0$. Predict only within the data range — extrapolation beyond it is unreliable.
The least-squares line
The fitted line predicts y for any x within the data.
For y-hat = 55 + 4x, predict the score when x = 3 hours.
55 + 4(3) = 55 + 12 = 67.
In y-hat = 55 + 4x, what does the slope 4 mean?
Slope = predicted change in y per 1-unit increase in x.
Using the model to predict far outside the observed range of x is called extrapolation, and it is unreliable.
Extrapolation is risky — the pattern may not continue.
In y-hat = a + bx, the letter a is the y-___.
a is the y-intercept: predicted y when x = 0.
Order the steps to predict a response from a regression equation.
Guard against extrapolation first, then compute and interpret.