Frameworks · Measuring performance

Simple linear regression

Fitting the straight line that best describes how one variable changes with another, with a scatter diagram, the correlation and, with care, extrapolation.

An IB toolkit tool, guide p. 15 HL only IB DiplomaUndergraduateMBAProfessional

Section 01: The framework at a glance

Simple linear regressionA scatter diagram of eight months of fictional data: advertising spend from $2,000 to $9,000 a month against sales revenue from $21,000 to $40,000. The least-squares line of best fit is y = 15.64 + 2.68x, in thousands of dollars, and the correlation coefficient is 0.97, a strong positive correlation. A dashed extension extrapolates to a spend of $11,000, predicting sales of about $45.1k. Correlation is not causation, and extrapolation assumes the relationship holds beyond the data.01020304050024681012Advertising spend ($000 a month)Sales revenue ($000 a month)Extrapolated: 45.1Line of best fit: y = 15.64 + 2.68xCorrelation coefficient r = 0.97: strong and positiveCorrelation is not causation. Extrapolating beyond the data assumes the relationship still holds there, which may notbe true.Fictional data, eight months.
Simple linear regressionA scatter diagram of eight months of fictional data: advertising spend from $2,000 to $9,000 a month against sales revenue from $21,000 to $40,000. The least-squares line of best fit is y = 15.64 + 2.68x, in thousands of dollars, and the correlation coefficient is 0.97, a strong positive correlation. A dashed extension extrapolates to a spend of $11,000, predicting sales of about $45.1k. Correlation is not causation, and extrapolation assumes the relationship holds beyond the data.0255004812Sales revenue, $000 a monthAdvertising, $000 a monthLine of best fit: y = 15.64 + 2.68xr = 0.97: a strong positivecorrelationExtrapolated to 11: about 45.1 (theopen circle)Correlation is not causation, andextrapolating assumes the patternholds beyond the data. Fictional data.

Simple linear regression HL A scatter diagram, the line of best fit, correlation and extrapolation.
Text description

A scatter diagram of eight months of fictional data: advertising spend from $2,000 to $9,000 a month against sales revenue from $21,000 to $40,000. The least-squares line of best fit is y = 15.64 + 2.68x, in thousands of dollars, and the correlation coefficient is 0.97, a strong positive correlation. A dashed extension extrapolates to a spend of $11,000, predicting sales of about $45.1k. Correlation is not causation, and extrapolation assumes the relationship holds beyond the data.

Section 02: What it is, and what it is for

The guide's tool includes scatter diagrams, the line of best fit, correlation and extrapolation. A scatter diagram plots pairs of values, such as advertising spend and sales, one on each axis. The line of best fit, found by least squares, is the straight line y = a + bx that makes the squared vertical distances from the points to the line as small as possible; b, the gradient, shows how much y changes for each extra unit of x. The correlation coefficient r, from −1 to +1, shows the strength and direction of the linear relationship.

Extrapolation extends the line beyond the data to forecast, which assumes the relationship holds outside the range observed. Correlation shows that two variables move together, not that one causes the other.

What it is for

To test whether two business variables are linked and to make cautious forecasts, such as sales from advertising spend or demand from temperature. The toolkit lists it as an HL tool for units 1, 4 and 5; its full guide is in preparation.

Section 03: How it is built

The parts of the framework, and what each one holds.

The two variables
The independent variable, x, which is thought to influence the other, and the dependent variable, y, such as advertising spend and sales revenue.
The scatter diagram
Each observation plotted as a point, which shows the direction, shape and strength of any relationship, and any outliers.
The line of best fit
The straight line y = a + bx that fits the points most closely. The gradient, b, is the change in y for each unit of x; the intercept, a, is the value of y when x is zero.
Correlation
How closely the points follow the line, from +1 (a perfect positive relationship) through 0 (none) to −1 (a perfect negative one).
Interpolation and extrapolation
Predicting y within the range of the data (interpolation) or beyond it (extrapolation), which assumes the relationship continues.

Section 04: When to use it

The business situations it suits, and where another tool will serve better.

It suits

  • Testing whether two business variables move together, such as advertising and sales, or temperature and ice cream demand, before planning around the link.
  • Making a short-range forecast within or just beyond the observed data.
  • Estimating how much one variable changes for each unit of another, such as extra revenue per extra thousand spent.

Look elsewhere when

  • Proving that one variable causes another; a third factor may drive both.
  • Relationships that curve, such as advertising with diminishing returns, or forecasts far beyond the data.

Section 05: How to use it, step by step

A short method. The toolkit guide gives the full one, with the evidence each step needs.

  1. Choose the two variables and decide which is the independent variable, x.
  2. Plot the data on a scatter diagram and look at the pattern.
  3. Fit the line of best fit and read its gradient in business terms.
  4. Judge the strength of the relationship from the correlation.
  5. If you extrapolate, stay close to the data and state the assumption that the relationship continues.

Section 06: Applied: Willow and Oak tests its advertising

Fictional business

Willow and Oak, a fictional garden furniture maker, plots eight months of data, as in the diagram above: advertising spend from $2,000 to $9,000 a month against sales revenue from $21,000 to $40,000. The line of best fit is y = 15.64 + 2.68x, in thousands of dollars, and the correlation coefficient is 0.97, a strong positive relationship.

Each extra $1,000 of advertising is associated with about $2,680 of extra sales revenue. Extrapolated to a spend of $11,000, beyond any month observed, the line predicts sales of about $45,100.

What it showed

The link is strong, but the firm treats it with care: sales also rise in spring, when it advertises most, so the season may explain part of the pattern. It raises spending gradually and checks each month's result, rather than trusting the extrapolated figure.

Section 07: Reading the result

How to interpret what it shows, and what that does and does not justify deciding.

Read three things in order: the strength of the relationship (the correlation), its size (the gradient, in business terms, such as dollars of sales per dollar of advertising), and its limits (the range of the data and the possible third factors).

A strong correlation shows that the two variables move together in this data, not that one causes the other. Before acting on the gradient, ask what else changed at the same time, such as the season, and whether the relationship is likely to hold outside the range observed. The further a prediction lies beyond the data, the less it should be trusted.

What it can justify deciding

  • Planning on a relationship that is strong, plausible and checked for third factors, within the range of the data.
  • Testing the relationship further with a small, controlled change, such as raising spending in one region first.

What it does not justify on its own

  • Concluding that changing x will change y, without other evidence of cause.
  • Committing large sums on an extrapolated figure far beyond the observed data.

Section 08: Strengths and limits

Why it helps

  • It turns a pattern in data into a measured relationship that can be compared and tested.
  • The gradient gives a figure managers can use in planning.
  • Correlation summarizes the strength of the link in one number.

Where it falls short

  • Correlation does not show causation; a third factor may drive both variables.
  • Extrapolation beyond the data assumes the relationship continues, which may not hold.
  • It captures straight-line relationships only, and a few outliers can distort the line.

Section 09: Common mistakes

Errors often made with it, by students and by managers, and how to avoid each.

  1. Mistake Reading correlation as causation.

    Instead Name a possible third factor and say how the business could test the link, such as a controlled trial.

  2. Mistake Extrapolating far beyond the data.

    Instead State the range of the data and treat predictions beyond it as assumptions, not findings.

  3. Mistake Giving the gradient without its meaning.

    Instead Translate it into business terms, such as 'each extra $1,000 of advertising is associated with about $2,680 of extra revenue'.

Section 10: Apply it

The application mode: the platform's tools that apply this framework, and a worksheet for your own case.

Frameworks structure judgement; they do not produce answers. Each one organizes evidence so that you can see it, test it and argue from it, and the decision, and the reasons for it, remain yours.

No interactive tool on the platform applies Simple linear regression in full yet, so the worksheet below is its application mode, with the tools that help with parts of it.

For part of it

  • Market Studio Marketing budget allocation See diminishing returns made explicit: the tool's response curves flatten as spending rises, a reminder that a straight line fitted to past spending cannot be extended far beyond the data.

Your worksheet

Work through the framework part by part for a business you know, then say what the result does and does not justify.

Kept on this device as you type, and nowhere else. Not saved without JavaScript: print the page or copy your answers before you leave.

For example: How closely are these two things linked, and what might happen next?
Who it is, what it sells, to whom, and what has changed.
Part by part

For each part, write what you find for this business and the evidence for it. Leave a part blank only if you can say why it does not apply.

What the framework shows when the parts are read together, and how sure you are of it.
The decision the result supports, what else you would need to know, and what would change your mind.

The Framework Application Lab: choose a framework by the decision you face.

Section 11: Try it

The platform's own exercise. Write your answer, then open the guidance.

An ice-cream kiosk believes its daily sales rise with the temperature. Describe how it should use simple linear regression to test this belief and forecast sales, and give two reasons to be cautious about a forecast for a day hotter than any it has recorded.

Kept on this device as you type, and nowhere else. Your answer is not saved without JavaScript; copy it before you leave.

Guidance: what a strong answer does
  • Name the dependent and independent variables and the data the kiosk would need.
  • Explain what the gradient and the correlation coefficient would tell it.
  • For caution, consider extrapolation and other factors such as weekends or school holidays.
  • Suggest how the kiosk could check the forecast in practice.

In the glossary

  • Simple linear regression A statistical method that fits a straight line to data on two variables, an independent variable (x) and a dependent variable (y), to describe how y tends to change as x changes and to make forecasts.
  • Scatter diagram A graph that plots pairs of values for two variables as points, with the independent variable on the horizontal (x) axis and the dependent variable on the vertical (y) axis, to show whether they are related.
  • Line of best fit The straight line that most closely follows the pattern of points on a scatter diagram, with the points balanced as evenly as possible above and below it.
  • Correlation The extent to which two variables are linearly related: with positive correlation they tend to rise together, with negative correlation one tends to fall as the other rises, and with no correlation there is no straight-line pattern.
  • Extrapolation Projecting a trend or a line of best fit beyond the data already collected, for example into future periods, to produce a forecast.
  • Sales forecast An estimate of a business's future sales, in units or in value, over a stated period, based on past data, research or judgement.

Read with

Compare it with any other framework in the library.

The explanation, method, example and exercise are the platform's own. Statements about the IB course cite the Business management guide; the business in the example is fictional.