Carrying Out a Test for the Slope (Extension)
IntroducidoThe test statistic
- Slope inference is extension content beyond the revised AP exam. The slope test statistic compares $b$ to $0$ in standard errors:
-
$$t = \frac{b}{SE_b}$$
- Both $b$ and $SE_b$ come straight from the computer output.
- Output often prints this $t$ (and its p-value) for you already.
Find the p-value
- Use the $t$-distribution with $df = n - 2$.
- Two-sided $H_a$: the standard output p-value (already two-tailed).
- One-sided $H_a$: halve the standard two-sided p-value when $b$ is in the claimed direction; otherwise use $1-p_{two}/2$. Confirm which alternative the software reports.
- The p-value is a tail probability at least as extreme in the direction(s) specified by $H_a$, assuming $\beta=0$ and the regression model.

The t statistic for the slope and its two-sided p-value
Make a decision
- Compare the p-value to $\alpha$:
- p $\le \alpha$ → reject $H_0$: convincing evidence of a linear relationship.
- p $> \alpha$ → fail to reject: not convincing evidence.
- Standard decision rule.
Conclude in context
- State it about the linear relationship between the real variables.
- "Convincing evidence of a positive linear relationship between hours and score" (if rejected).
- Or "not convincing evidence of a linear relationship" (if not).
- Failing to reject never proves $\beta = 0$.
Computer output's p-value is two-sided — halve it for a one-sided $H_a$ (and only if the sample slope points the way $H_a$ claims). Use $df = n - 2$, and read $t = b/SE_b$ from the output rather than recomputing. As always, "fail to reject" means insufficient evidence, not proof of no relationship.
Output: $b = 4.2$, $SE_b = 1.0$, $n=20$; two-sided p-value $\approx 0.000538$; $H_a: \beta > 0$, $\alpha = 0.05$.
- $t = \dfrac{4.2}{1.0} = 4.2$ ($df = 20-2=18$).
- One-sided p-value: $0.000538 / 2 \approx 0.000269$ ($b > 0$ matches $H_a$).
- Decide: $0.000269 \le 0.05$ → reject $H_0$; convincing evidence of a positive linear relationship.
Compute the slope $t = \frac{b}{SE_b}$ from the output, find the p-value at $df = n - 2$ (for a one-sided $H_a$, use half the two-sided value if the observed slope matches its direction, otherwise its complement), and compare to $\alpha$: reject $H_0$ if p $\le \alpha$ (a linear relationship exists), else fail to reject — stated in context.
Write from the data: find the test statistic
- The fitted slope is 1.5.
- Its standard error is 0.5; test the null $\beta=0$.
For H0 β = 0, the fitted slope is b = 1.5 and its standard error is SE_b = 0.5. What is the t statistic?
t for a slope is the estimate divided by its standard error.
For H0 β = 0, the fitted slope is b = 1.5 and its standard error is SE_b = 0.5. What is the t statistic?
Dificultad 1/5 · 100 monedas + un pequeño bono en el primer intento de hoy
1.5 divided by 0.5 is 3. Other options — 4.2 (the other regression); 1.5 (the slope).
Testing the fitted slope against zero
t = b/SE_b measures how far the slope sits from 0.
b = 4.2, SE_b = 1.0. Compute the slope test statistic t = b/SE_b.
Dificultad 2/5 · 140 monedas + un pequeño bono en el primer intento de hoy
4.2 / 1.0 = 4.2.
A two-sided output p-value is 0.001. For a one-sided Ha (with b in the claimed direction), what is the p-value?
Dificultad 2/5 · 140 monedas + un pequeño bono en el primer intento de hoy
Halve the two-sided value: 0.001 / 2 = 0.0005.
With a p-value of 0.0005 and α = 0.05, the decision is...
Dificultad 1/5 · 100 monedas + un pequeño bono en el primer intento de hoy
0.0005 ≤ 0.05 → reject H0.
Standard regression coefficient output commonly reports a two-sided p-value for testing the slope against zero; confirm the software’s stated alternative.
Dificultad 1/5 · 100 monedas + un pequeño bono en el primer intento de hoy
For the standard symmetric t test, use p-two/2 only when the observed slope matches a preplanned one-sided alternative; use 1 − p-two/2 in the opposite direction.
Rejecting H0: β = 0 lets you conclude...
Dificultad 1/5 · 100 monedas + un pequeño bono en el primer intento de hoy
Reject the flat-line null → a real linear relationship.