QBUS5001 Chap.10 Simple Linear Regression and Correlation
Simple Linear Regression and Correlation
This chapter begins with a scatterplot, covariance and correlation, then assigns predictor and outcome roles in a simple fitted line. Slope, intercept, fitted values and residuals are interpreted in business units. Correlation remains distinct from slope and causation. Prediction is audited for extrapolation, residual patterns and omitted mechanisms, with the current source boundary stated for later regression material.
The scatterplot establishes direction, form, strength, clusters and influential observations before a line compresses that pattern. Covariance records joint direction in combined units, while correlation standardises linear association to a scale from minus one to one.
A fitted slope reports the estimated change in the outcome for a one-unit increase in the predictor; the intercept reports the fitted outcome at a zero predictor and may be meaningless when zero lies outside the observed range. A fitted value sits on the line, and a residual is observed minus fitted, so its sign must match the point's position above or below the line. Prediction outside the data range is extrapolation.
Even a strong fitted association remains vulnerable to omitted variables and cannot establish causation without an appropriate design.
What this chapter covers
- 01
Scatterplot form, direction, strength and unusual points
- 02
Sample covariance and measurement units
- 03
Sample correlation from minus one to plus one
- 04
Outcome, predictor, slope and intercept
- 05
Fitted values and observed residuals
- 06
Prediction range and the causation boundary
Worked example · free
Interpret a fitted line and residual
- +1Interpret the slope as 2.4 thousand dollars of higher predicted sales per additional thousand dollars of advertising.
- +1Discuss the intercept only if zero advertising belongs to a meaningful evidence range.
- +1Calculate predicted sales 18 plus 2.4 times 6 equals 32.4 thousand dollars.
- +1Calculate residual observed minus predicted: 35 minus 32.4 equals 2.6 thousand dollars.
- +1State that the positive residual means the line underpredicted this observation and avoid a causal claim.
Key terms
- Regression Slope
- The predicted change in mean outcome for a one-unit increase in the predictor.
- Regression Intercept
- The predicted outcome when the predictor equals zero, meaningful only when zero is relevant to the data.
- Fitted Value
- The outcome predicted by the fitted regression equation for a stated predictor value.
- Regression Residual
- The observed outcome minus its fitted value.
- Extrapolation
- Prediction outside the predictor range represented in the fitted data.
Simple Linear Regression and Correlation FAQ
How are correlation and slope different?
Correlation is unitless, symmetric in the two variables and bounded from minus one to plus one. Slope assigns predictor and outcome roles and has outcome units per predictor unit. Rescaling a variable changes slope but not correlation.
What does a positive residual mean?
Because residual equals observed outcome minus predicted outcome, a positive residual means the observation lies above the fitted line and the model underpredicted it. A negative residual means the model overpredicted the observed outcome.
Can a regression coefficient be interpreted causally?
Not from an observational fitted equation alone. Confounding, selection, common trends or reverse direction may explain the association. Causal interpretation requires a design and assumptions that isolate the effect of changing the predictor.
Why is extrapolation risky?
The fitted relationship is supported by predictor values present in the data. Outside that range, the line may no longer describe the process and can generate unrealistic predictions. State the observed range and gather comparable evidence before extending it.
Exam move
Sketch the scatterplot and label predictor and outcome before calculating anything. Interpret covariance, correlation and slope in separate sentences so their units and roles stay distinct. For each fitted value, compute the residual as observed minus predicted and verify the sign visually.
Audit every prediction for evidence range, residual pattern and a plausible business mechanism, then limit causal wording to what the design supports. Practise from a graph before using an equation. Mark two points on a fitted line and compute rise over run to recover the slope in original units.
Evaluate the line at one observed predictor value, subtract the prediction from the observed outcome and check that the residual sign agrees with the plot. Change the measurement unit of the predictor and explain why the slope changes while correlation does not. Interpret the intercept only after checking whether zero is within a meaningful range.
For prediction questions, draw the observed predictor interval and flag any requested value outside it. Finish with a two-sentence audit: one sentence states the fitted association and forecast, while the second names a residual pattern, omitted mechanism or design limitation that prevents a causal conclusion. Create a final interpretation table with separate rows for correlation, slope, intercept, fitted value and residual.
Give each row its units and a sentence pattern. Then add a boundary column for linearity, range or causation. Keeping these objects separate prevents a strong correlation from being described as a large slope and stops a useful prediction from becoming an unsupported causal claim.
Working through Simple Linear Regression and Correlation in QBUS5001? Sia is AskSia’s AI Statistics tutor — ask any QBUS5001 Simple Linear Regression and Correlation question and get a clear, step-by-step explanation grounded in how QBUS5001 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.