Science · Scientific investigation
Predicting Relationships and Outcomes
Estimated time includes reading and one quiz. Take the time you need.
Start here: key ideas
- In a direct relationship, two variables move in the same direction.
- In an inverse relationship, one rises as the other falls.
- Interpolation predicts within the measured range.
- Extrapolation predicts beyond the measured range and is less certain.
- Correlation alone does not prove causation.
What you’ll be able to do
- Identify direct, inverse, or no clear relationship.
- Make an interpolation from data.
- Explain why extrapolation is less certain.
- Tell correlation from causation.
Recognize common relationships, make careful predictions, and avoid confusing correlation with cause.
Optional review
If you’d like to review the basics, visit Data, Evidence, and Scientific Explanations. You can start this lesson without completing those first.
Direct, inverse, or absent
In everyday descriptions, a positive relationship means variables tend to increase together; a negative relationship means one tends to fall as the other rises. Some science questions call these direct and inverse trends. In mathematics, direct proportion is stricter: y/x is constant; inverse proportion requires xy constant. A trend alone does not establish either formula.
Read both axes, units and the measured interval. Describe the overall trend and important exceptions. A curve that rises and then levels off does not support unlimited straight-line extrapolation.
Make a bounded prediction from a graph
Follow one named substance’s curve and keep track of the units. A prediction between shown values is interpolation. A prediction beyond the shown range is extrapolation and needs more caution. Do not assume every substance follows the same curve.
Make a bounded prediction from a graph — OpenStax College, Chemistry (CC BY archive). CC BY 3.0. Original source image reproduced unchanged; explanatory caption added.
Predict within the evidence
Interpolation predicts a value inside the measured range. It is usually safer because nearby data support it. Extrapolation predicts beyond the measured range and assumes the pattern continues.
Example: data measured from 10°C to 30°C can support an estimate at 20°C better than one at 80°C. State uncertainty when making any prediction.
Worked example. A graph shows heart rate at workloads 1, 2, and 3 as 80, 100, and 120 beats per minute. A value near 110 at workload 2.5 is an interpolation. Predicting workload 10 from these three points is an uncertain extrapolation.
Worked table estimate: rates of 12 and 20 events/min at 10 and 20 °C suggest about 16 events/min at 15 °C if a near-linear trend is reasonable between measurements. This is an estimate, not a measured value. A plateau or reversal is evidence to revise the model, not discard inconvenient data.
Control changes before claiming cause
Correlation means two variables are associated. It does not prove that one caused the other. A third variable, chance, or reverse direction may explain the pattern.
A controlled experiment gives stronger evidence of causation. Change one main factor, hold other conditions steady, and compare the measured result.
To test cause, change the suspected cause while holding other important factors steady. If hotter days and ice-cream sales occur together, heat may affect sales, but the correlation does not show that ice cream causes hot weather.
Terms to remember
- direct relationship
- A pattern in which two variables move in the same direction.
- inverse relationship
- A pattern in which one variable rises as the other falls.
- interpolation
- A prediction inside the measured data range.
- extrapolation
- A prediction beyond the measured data range.
- correlation
- A measured association between variables.
- causation
- A relationship in which one change produces another.
Your quick summary
- Direct variables move together.
- Inverse variables move in opposite directions.
- Interpolation stays inside the data range.
- Extrapolation goes beyond the data range.
- Correlation does not by itself prove causation.
Lesson quiz
Check what you learned, then review the explanations. This quiz saves results on this device only. Sign in to save quiz results to your account.