01 · Learn the idea
Three fictional calibration tables
Use t=0, 1, 2, 3, 4. Table A has y=3, 5, 7, 9, 11; table B has y=1, 2, 5, 10, 17; table C has y=2, 4, 8, 16, 32. First calculate first differences, second differences, and successive ratios. Constant first differences support a linear model; constant second differences support a quadratic at equally spaced inputs; constant ratios support an exponential for these positive outputs.
Enter all three tables into a regression-capable tool. Obtain a linear, quadratic, and exponential regression for each table, saving equations, residuals, and plots. Record the tool’s exponential convention: a·b^t versus a·e^(kt). Do not compare coefficients from different conventions without translating.
02 · Build understanding
Prediction and model limitations
The exact natural models are A:2t+3, B:t²+1, C:2·2^t. They predict 13, 26, 64 at t=5. A quadratic fit can also represent a line with zero quadratic coefficient; prefer the simpler adequate model. Inspect residual patterns as well as a summary fit statistic.
These are constructed data, not evidence that a real device continues forever. Discuss how an out-of-range prediction at t=50 may fail because of saturation, changing mechanisms, or an inappropriate input domain. Design one additional measurement that would help distinguish two plausible models.
Apply and explain
Your turn
Submit nine technology-generated fits, a selection argument for each table, residual plots, the t=5 predictions, and an extrapolation critique.
Use paper or explain your thinking aloud before opening the solution.
Check your reasoning
The tables were generated by the stated natural models, but fit alone does not establish a physical mechanism. Repeated exact agreement is a property of these fictional tables.
- Differences/ratios
- Nine technology fits
- Residual evidence
- Predictions
- Extrapolation limits
Ready to practice?
Try the questions, get feedback, and review your reasoning.
Compare linear, quadratic, and exponential fits practice →