Study guide: Bivariate Trends, Predictions, and Residuals
Paired measurements describe how two variables vary together. A trend line summarizes association, predicted outputs can be compared with observations using residuals, and the observed input range limits how confidently that trend can be extended.
Ideas to explain in your own words
- A scatter representation plots each input with its paired output. A rising pattern indicates positive association in the supplied record, whereas a straight pattern adds a claim about linear structure. Association does not establish causation.
- A trend equation computes a predicted output from a supplied input. Substitution gives the model’s value, which may differ from any actual measurement even when the trend is useful. Do not label a predicted value as a measurement.
- A residual is observed output minus predicted output. Its sign identifies whether an observed point sits above or below the model, and its magnitude measures that vertical discrepancy. Subtract predicted from observed in that order.
- Interpolation predicts within measured input bounds. Extrapolation predicts outside those bounds, where a fitted pattern may fail because the data did not constrain that part of the model. An equation can produce an extrapolated value without supporting its reliability.
Useful vocabulary
- association
- A pattern connecting variation in paired variables.
- residual
- Observed value minus predicted value.
- extrapolation
- Prediction outside the range of observed inputs.
Watch your reasoning
- Association does not establish causation. The supplied exact pattern is small and artificial; describe its direction and form without generalizing to unobserved people, future measurements or a causal mechanism.
- Do not label a predicted value as a measurement. Preserve the variable units and any stated model range; real observations may vary around the trend because a line summarizes rather than captures every factor.
- Subtract predicted from observed in that order. A negative residual means overprediction, not a negative observation; compare residuals in the output’s units rather than treating them as input-coordinate differences.
A worked review example
Complete supplied guide examples: Describe a Supplied Paired Record — A supplied scatter record has pairs (1,2), (2,4), (3,6), (4,8). Which description fits? Candidate answers: (A) Negative linear association (B) No change in y as x changes (C) Positive linear association in this record (D) The record proves x causes y. Each increase of 1 in x accompanies an increase of 2 in y; the four points lie on a rising line. Association in the record does not identify a cause. Distinguish a Model Prediction from an Observation — A supplied trend model is y=3x+2. What y does it predict at x=3? Candidate answers: (A) 10 (B) 11 (C) 12 (D) 22. Substitute x=3: y=3(3)+2=11. This is the line model’s prediction, not a promised observation. Measure Signed Prediction Error — At x=2, model y=2x+1 predicts 5, but the observed y is 3. Find observed−predicted. Candidate answers: (A) -2 (B) 0 (C) 2 (D) 8. Residual=observed−predicted=3−5=−2. The observed point is 2 output units below the line. Notice a Prediction Outside the Data Range — A model was fitted to measurements at x=2 through x=10. Which proposed prediction is extrapolation? Candidate answers: (A) Predicting at x=14 (B) Predicting at x=2 (C) Predicting at x=3 (D) Predicting at x=8. 14 lies outside the measured interval 2 through 10. Extending a trend outside its data range is extrapolation and requires more caution than interpolation.
Make a practice plan
- Study the four methods in the listed order; attempt each independent check before opening its explanation.
- Complete each lesson’s three changed-input questions and record the operation or reasoning step behind any correction.
- Take the four-question unit review, then solve the transfer task with written justifications.
- After correcting errors, attempt the separate unit exam and revisit the linked lesson for any unresolved method.