Toolkit
Simple linear regression
A method of fitting a straight line to data on two variables, used to describe a relationship and to forecast.
Section 01: What it is for
To measure how closely two variables move together, describe the relationship with a straight line, and make cautious forecasts, such as how sales respond to price or to advertising.
- Class
- Planning and Situational (the platform's reading; the TSM does not classify this tool)
- Suggested units
- Unit 1, Unit 4, Unit 5 (guide, p. 14)
- The guide includes
- Scatter diagrams; Line of best fit; Correlation and extrapolation
When to use it
- When there is reason to think that one variable affects another, and paired data are available.
- When a forecast is needed and a straight line fits the data reasonably well.
- When testing a claim, such as that advertising raises sales.
How to build it
- Choose the independent variable (x) and the dependent variable (y), and collect pairs of values.
- Plot a scatter diagram, and look for a pattern and for outliers.
- Fit the line of best fit, y = a + bx, by least squares or with software, and read the gradient b in business terms.
- Judge the strength and direction of the relationship from the correlation coefficient r, which runs from −1 to +1.
- Forecast within the range of the data where possible; if you extrapolate, say so and state the assumption that the relationship continues.
The evidence it needs
- Paired observations of the two variables, ideally from what people actually did.
- Enough observations, gathered under similar conditions.
Section 02: What it shows, and what it hides
- The direction and strength of a straight-line relationship.
- How much y changes, on average, for each unit change in x.
- Whether x causes y: a third factor may drive both.
- Whether the relationship holds beyond the range of the data.
Limitations
- Few or unreliable observations give an unreliable line.
- It fits straight lines only, so a curved relationship is misdescribed.
- Outliers can pull the line, and inflate or hide the correlation.
Section 03: Applied: Atelier Barro's prices and sales
From the case Atelier Barro.
Exhibit 3 estimates how many mugs a month Atelier Barro would sell at five prices: 700 at €20, 640 at €22, 590 at €24, 560 at €25 and 480 at €27. The least-squares line of best fit is mugs a month = 1,315 − 30.5 × price, with a correlation coefficient of −0.995, a very strong negative relationship: each €1 on the price is associated with about 30 fewer mugs a month. At €23, between two observed prices, the line predicts about 612 mugs.
Section 04: In the exam
At HL, simple linear regression is a toolkit tool for units 1, 4 and 5, covering scatter diagrams, the line of best fit, and correlation and extrapolation. Questions may ask for a scatter diagram or a line to be interpreted, for a forecast, or for the reliability of a forecast to be evaluated. Comment on the strength of the correlation, the risk of extrapolation and the difference between correlation and causation.
Common errors
- Claiming that a strong correlation proves cause and effect.
- Extrapolating far beyond the data without warning that the relationship may not hold.
- Reading the gradient without its units, such as mugs a month for each euro.
Arjun Agrawal | The Business Brief, https://the-business-brief-dun.vercel.app/toolkit/simple-linear-regression/. Original teaching material, not an IB publication. The organizations in the teaching cases are fictional unless a case says otherwise.
The relationship looks almost perfect, but the evidence is weaker than the correlation suggests: three of the five points come only from a survey of what customers say they would do, one from a survey and a two-week price test, and only one from actual sales. Extrapolating to €30 would predict about 398 mugs, beyond any price the studio has tried. The line is a reasonable guide between €20 and €27, and Inês would be wise to test any new price before relying on it.