With seven models on offer, the question isn't can a regression calculator fit your data — it's which model you should trust. The answer comes from data shape, theory, and a careful look at r².
Start with the shape
Plot the points (even roughly) and match the pattern:
| If the data… | Try | Model |
|---|---|---|
| Follows a straight line | Linear | y = A + Bx |
| Rises then falls (or a single bend) | Quadratic | y = A + Bx + Cx² |
| Climbs fast then flattens | Logarithmic | y = A + B·ln x |
| Grows by a constant percentage | Exponential | y = A·e^(Bx) or A·Bˣ |
| Follows a scaling law | Power | y = A·xᴮ |
| Falls toward an asymptote | Inverse | y = A + B/x |
Shortlist and fit
Pick the two or three shapes that look plausible and let domain knowledge guide you — radioactive decay is exponential, the area of a circle vs radius is power-law. Then fit each in STAT mode and note the coefficients and r.
Compare with r²
r² is the share of variation the model explains, from 0 to 1. Between candidates of similar complexity, the higher r² usually wins.
Fit the same data with two models and compare directly. The mechanics of r and r² are covered in the statistics calculator pillar guide, and the linear/curve distinction in linear vs nonlinear regression.
The overfitting trap
A more complex model (e.g. quadratic over linear) can only increase r², even when the extra term is just chasing noise. Prefer the simplest model that fits well and extrapolates sensibly. If two models tie, pick the one your theory supports.
Frequently asked questions
How do I choose a regression model?
Read the scatter shape, fit two or three candidates, and compare r² — favouring the simplest good fit.
Does higher r² always win?
No — added complexity can inflate r². Balance fit with simplicity and theory.
What if two models tie?
Choose the simpler one or the one your domain knowledge supports.
Compare models on your data
Fit several regression types in STAT mode and let r² guide the choice.
Open the regression calculator →