Glossary
Mathematical Modeling
What is mathematical modeling?
Mathematical modeling, also spelled mathematical modelling, expresses a proposed account of a system in equations. In behavioral science, a model might describe how people learn from feedback or choose between options. Making the account precise lets researchers derive predictions and compare them with data.
Types of mathematical models
- Deterministic: the same inputs and starting conditions produce the same output.
- Stochastic: the model includes randomness and describes probabilities of outcomes.
- Static: the model describes relationships without modeling how they unfold over time.
- Dynamic: the model describes how a system changes over time.
These distinctions overlap. A model can be both dynamic and stochastic.
A small model-building example
Suppose a researcher wants to predict scores on a ten-question quiz from the number of practice questions a student completed. A simple starting model is predicted score = a + b × practice questions. Here, a is the predicted score with no practice questions, and b is the predicted score difference associated with one additional practice question.
The researcher estimates these parameters using one set of students, then checks the predictions against scores from students held out of that fitting process. Suppose the fitted values are a = 3 and b = 0.2: ten practice questions would predict a score of five. These numbers are illustrative, not a measured effect of practice.
The equation assumes a straight-line relationship. At forty practice questions it predicts eleven correct answers, which is impossible on a ten-question quiz. That failure reveals a limit: the model may need a bounded outcome or a narrower range of use. Prior knowledge could also influence both practice and scores, so even accurate predictions would not show that assigning one more practice question causes the predicted improvement.
What a good fit tells you
A model that matches observations has passed a useful test. It has not proved the theory behind it. Different models can predict similar results, and a model that fits better than its competitors can still fit poorly.
Researchers can use simulated data to check whether their analysis recovers known parameters and distinguishes competing models. They can also check whether the fitted model reproduces important patterns and predicts observations it was not fitted to.
Using a model to change behavior
Prediction and intervention are different questions. Predicting who will complete a course does not, by itself, tell you how to make more people complete it. That requires a justified causal account and evidence about the proposed intervention.
Models can make assumptions explicit and help compare possible designs. Their usefulness depends on those assumptions, the measurements, and how well the model works outside the data used to build it.