AP Statistics (AP Stats): Checking Conditions (Linear, Independent, Normal, Equal Variance, Random – LINER)

AP Statistics – Checking Conditions (Linear, Independent, Normal, Equal Variance, Random – LINER)

AP Statistics Study Guide: Checking Conditions (LINER)

What This Is

Checking conditions (LINER) ensures that statistical methods (like confidence intervals, hypothesis tests, or regression) are valid. Without verifying these, your conclusions may be incorrect. For example, if a researcher wants to test whether a new fertilizer increases crop yield, they must confirm that the data meets Linear, Independent, Normal, Equal variance, and Random conditions before running a t-test or regression analysis.

Key Terms & Formulas

• LINER Conditions: The five assumptions for inference in regression and means:
• Linear: Relationship between variables is linear (check scatterplot/residual plot).
• Independent: Observations are independent (check 10% condition if sampling without replacement).
• Normal: Sampling distribution of residuals is approximately normal (check histogram/Q-Q plot of residuals or n-30 for means).
• Equal Variance: Residuals have constant spread (check residual plot for no "fanning").

• Random: Data comes from a random sample or randomized experiment.

• 10% Condition: For independence, if sampling without replacement, n-0.10N (sample size-10% of population).

• Normal Condition for Means: If n < 30, check for normality (histogram, boxplot, or normal probability plot). If n-30, Central Limit Theorem (CLT) applies.

• Normal Condition for Proportions: np-10 and n(1-p)-10 (for confidence intervals, use p?; for tests, use p?).

• Residual Plot: Plot residuals (y – ?) vs. x or-to check Linear and Equal variance conditions.

• Q-Q Plot (Normal Probability Plot): Used to check Normal condition; points should follow a straight line.

• Hypothesis Test for Slope (?): H?:-= 0 (no linear relationship), H?:-? 0 (or one-sided).

• t-test for Slope: t = (b – ) / SEb, where b = sample slope, SEb = standard error of slope (from LinRegTTest on TI-84).

• TI-84 Commands:

• LinRegTTest: Performs linear regression t-test (checks slope significance).
• RESID: Stores residuals (use 2nd-STAT-RESID after running LinReg(a+bx)).
• invT(area, df): Finds critical t-value (e.g., invT(0.975, 18) for 95% CI with df=18).

Step-by-Step / Process Flow

For Regression Inference (FRQ Example)

1. State Hypotheses:
2. H?:-= 0 (no linear relationship between x and y)

3. H?:-? 0 (linear relationship exists)

4. Check LINER Conditions:

5. Linear: Scatterplot shows linear pattern; residual plot has no curve.
6. Independent: 10% condition (if sampling without replacement) or randomized experiment.
7. Normal: Histogram/Q-Q plot of residuals is roughly normal (or n-30).
8. Equal Variance: Residual plot shows no "fanning" (constant spread).

9. Random: Data comes from a random sample or randomized experiment.

10. Compute Test Statistic:

11. Run LinRegTTest on TI-84 (L1, L2, Y1)-Record t and p-value.

12. Find p-value:

13. Compare p-value to? (usually 0.05).

14. Make Conclusion in Context:

15. If p < ?, reject H?: "There is convincing evidence of a linear relationship between x and y."
16. If p-?, fail to reject H?: "There is not convincing evidence of a linear relationship."

Common Mistakes

• Mistake: Forgetting to check the 10% condition for independence. Correction: Always verify n-0.10N when sampling without replacement, even if the problem doesn’t mention it.

• Mistake: Assuming normality for small samples without checking. Correction: For n < 30, plot residuals (histogram/Q-Q plot) to confirm normality.

• Mistake: Confusing "equal variance" with "no pattern" in residual plots. Correction: Equal variance means residuals have consistent spread (no "megaphone" shape), not necessarily no pattern.

• Mistake: Using p? instead of p? for the Normal condition in hypothesis tests. Correction: For tests, use p? (null value); for confidence intervals, use p?.

• Mistake: Ignoring the Random condition. Correction: Always state whether data comes from a random sample/experiment (critical for generalizability).

AP Exam Insights

• Tricky Distinction: The Normal condition for regression applies to residuals, not the original data.
• Common FRQ Setup: You’ll often be given a residual plot and asked to verify Linear and Equal variance conditions.
• Calculator Pitfall: LinRegTTest gives t and p-value for the slope, but you must still check LINER conditions manually.
• Z vs. T: Use t-procedures for means (unless-is known) and z-procedures for proportions.

Quick Check Questions

1. Multiple Choice: A researcher collects data on 25 randomly selected students to test if there’s a linear relationship between hours studied and exam scores. Which condition is not met for a linear regression t-test?
2. (A) Linear
3. (B) Independent
4. (C) Normal
5. (D) Equal Variance

6. (E) Random Answer: (C) Normal. With n = 25 < 30, we must check normality of residuals (not automatically met).

7. FRQ Part: A study examines the relationship between car age (years) and resale value ($). A residual plot shows a curved pattern. Which LINER condition is violated, and what does this imply? Answer: The Linear condition is violated. This implies a linear model is inappropriate; a nonlinear model may fit better.

Last-Minute Cram Sheet

1. LINER: Linear, Independent (10% condition), Normal (residuals for regression, CLT for means), Equal variance (residual plot), Random.
2. 10% Condition: n-0.10N for independence (sampling without replacement).
3. Normal for Means: n-30 (CLT) or check normality for small n.
4. Normal for Proportions: np-10 and n(1-p)-10 (use p? for CI, p? for tests).
5. Residual Plot: Check for Linear (no curve) and Equal variance (no fanning).
6. Q-Q Plot: Points should follow a straight line for Normal condition.
7. TI-84: LinRegTTest for slope test, RESID for residuals, invT(area, df) for critical values.
8. Always check 10% condition for independence, even if unstated.
9. Regression Normal condition applies to residuals, not original data.
10. Equal variance-no pattern (residuals can have random scatter but must have constant spread).