Skip to content
Tutorials Center
My learning

Topics

Lesson 34 of 37 · Functions and Modeling

Modeling with Functions

Choose and build linear, quadratic or exponential models from data using differences and ratios, then use them to make predictions, with code checks.

A mathematical model is a function that describes how some real quantity behaves. Once you have a good model you can make predictions: how much will it cost, how high will the ball go, when will the population reach a million?

The hard part is choosing the right kind of function. This lesson shows you a simple test using the pattern in the data.

Three families of models

Family Form Pattern in equally spaced data
Linear y = mx + b constant first differences
Quadratic y = ax² + bx + c constant second differences
Exponential y = a · bˣ constant ratios

To use the test, take the outputs for equally spaced inputs (such as x = 0, 1, 2, 3, ...) and compute differences or ratios between neighbors.

Example 1. Outputs: 5, 9, 13, 17, 21. First differences: 4, 4, 4, 4. Constant, so the model is linear with slope 4. Since y = 5 at x = 0, y = 4x + 5.

Example 2. Outputs: 1, 3, 7, 13, 21. First differences: 2, 4, 6, 8. These are not constant. Second differences: 2, 2, 2. Constant, so the model is quadratic with a = 2/2 = 1 (the second difference equals 2a). Try y = x² + bx + 1: at x = 1, 1 + b + 1 = 3, so b = 1. The model is y = x² + x + 1. Check x = 3: 9 + 3 + 1 = 13. Correct.

Example 3. Outputs: 2, 6, 18, 54. Ratios: 3, 3, 3. The model is exponential, y = 2 · 3ˣ.

Python Which family fits the data?

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

From data to a model

You can find a model in three steps:

  1. Look at the pattern and choose the family.
  2. Use known points to find the unknown numbers (m and b, or a, b, c, or a and b).
  3. Test the model on a data point you did not use.

Example 4: a ball in the air. A ball is thrown upward. Its height (meters) at 0, 1, 2, 3 seconds is 2, 17, 22, 17. Second differences: first differences are 15, 5, −5; second differences are −10, −10. The data are quadratic, with 2a = −10, so a = −5. Then h = −5t² + bt + 2. At t = 1: −5 + b + 2 = 17, so b = 20. The model is h(t) = −5t² + 20t + 2. Check t = 2: −20 + 40 + 2 = 22. Correct.

The ball's highest point is at t = −b/(2a) = 2 seconds, with height 22 m. Since the model has a vertex at t = 2, the ball is at its peak there, as the symmetric data suggested.

Python Fit a quadratic model and test it

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Using and questioning a model

A model gives predictions, but it is only as good as its assumptions.

  • Interpolation (predicting inside the range of your data) is usually safe.
  • Extrapolation (predicting far outside the data) can be badly wrong. A plant that grew exponentially for 5 weeks will not keep doing it forever.
  • Pay attention to the domain: a negative time or a negative number of people makes no sense.
  • Real data are rarely perfect. Small measurement errors mean differences are only roughly constant; you then choose the family that fits best.

Always state what the variables mean and their units, such as "h is the height in meters and t is the time in seconds". Without that, a model is just an equation.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. Outputs 3, 7, 11, 15 at x = 0, 1, 2, 3 suggest which model?

    The first differences are all 4, so it is linear: y = 4x + 3.
  2. 2. Outputs 4, 8, 16, 32 suggest which model?

    Each output is twice the one before, so y = 4 · 2ˣ.
  3. 3. If the second differences of equally spaced data are all 6, what is a in ax² + bx + c?

    The second difference equals 2a, so a = 3.
  4. 4. Why is extrapolating far beyond the data risky?

    A model fitted to one range need not describe a very different range.

Summary

  • Constant first differences mean linear, constant second differences mean quadratic, and constant ratios mean exponential.
  • Choose the family, solve for the unknown numbers using data points, and test with a point you did not use.
  • Interpolate with confidence; extrapolate with caution.
  • Always state what the variables and units mean.

Next: Practice Set: Polynomials and Factoring