Bivariate data, correlation and regression
Studying this course? Keep your progress in one place.
Key facts
- Bivariate data are pairs ; describe correlation by direction and strength, in context.
- A regression line : the gradient is the average change in per unit of .
- Interpolation (inside the data range) is reliable; extrapolation (outside) is not.
- : plot against (gradient , intercept ). : plot against (gradient , intercept ).
- Correlation does not imply causation.
Scatter diagrams and correlation
Describe the direction (positive, negative or none) and the strength (strong or weak), in context.
Bivariate data are pairs of values for each item. Correlation measures how closely the points follow a line. Strong means close to a line; weak means widely scattered.
A scatter diagram may show distinct sections of the population with different patterns. Then a single regression line may be inappropriate, and the groups should be considered separately.
Positive
- y tends to increase as x increases
Negative
- y tends to decrease as x increases
Zero
- No linear pattern
As the outside temperature rises, ice-cream sales rise and the points lie very close to a line. How should you describe this?
Explanatory and response variables
The explanatory variable goes on the horizontal axis and the response variable on the vertical axis.
The explanatory variable is the one you choose or think explains the other. The response variable is measured to see its response. A regression line of on has the form .
The gradient is the average change in for a one-unit increase in ; the intercept is the predicted when , which may have no meaning in context. For , height rises by cm per year of age on average.
- Regression line
- Gradient average change in per unit of
A regression line is with in cm and in years. Interpret the gradient.
Interpolation and extrapolation
Predictions inside the range of the data are reliable; predictions outside it are not.
Predicting inside the range is interpolation; predicting outside is extrapolation, which is unreliable because the relationship may not continue.
For ages 5 to 12, at gives cm (reliable), but at gives cm (unrealistic). Always give the reason: the value is outside the range of the data.
The data cover ages 5 to 12. Is a prediction at age 10 reliable?
Linear form of
Take logs: , so plot against .
For a power model , taking logs of both sides gives a straight line with gradient and vertical intercept .
Worked example
A plot of against is a straight line with gradient and intercept . Find when .
- 1
The gradient is , so .
- 2
The intercept is , so .
- 3
The model is ; at : .
For , what do you plot to get a straight line?
Linear form of
Take logs: , so plot against .
For an exponential model , the plot of against is a straight line with gradient and intercept . Use and .
Worked example
The line of against passes through and . Find and .
- 1
Intercept , so .
- 2
Gradient .
- 3
So and .
For , the plot of against has gradient . What is ?
Correlation and causation
No: correlation does not imply causation.
Two variables can be correlated because a third confounding variable affects both, by coincidence, or because one does cause the other. Only a controlled experiment, with random allocation of the explanatory variable, can establish causation.
When evaluating a claim, state the association found, say it does not prove causation, suggest another factor, and recommend how to test it.
- 1
Hot weather
more ice cream sold
- 2
Hot weather
more people swim, so more drownings
- 3
Correlation
ice cream sales and drownings rise together, but neither causes the other
| Third variable | Coincidence | Causation | |
|---|---|---|---|
| Reason | Affects both variables | No real link | One causes the other |
| Example | Hot weather raises ice-cream sales and swimming | Unrelated trends that happen to match | Needs a controlled experiment to show |
Third variable
- Reason:
- Affects both variables
- Example:
- Hot weather raises ice-cream sales and swimming
Coincidence
- Reason:
- No real link
- Example:
- Unrelated trends that happen to match
Causation
- Reason:
- One causes the other
- Example:
- Needs a controlled experiment to show
Ice-cream sales and drowning incidents are strongly positively correlated. What is the best explanation?
Try an exam question
A scientist models the mass grams of a plant against its age days, using data for to , by . Interpret the gradient, and comment on the reliability of using the model to predict the mass at .
[3 marks]
- [1]The mass is predicted to rise by g per day, on average.
- [1] is outside the range to , so this is extrapolation.
- [1]So the prediction is unreliable.
That's the notes covered.
Carry on to the next subtopic.
Try a question
What height does the model predict for a child aged 8 years?
1 mark, marked against the mark scheme
What height does the model predict for a child aged 8 years?
1 mark, marked against the mark scheme