FigureCheckFits and statistics
Interpret fits and statistical checks
Fit each visible series independently, inspect validation evidence, and compare groups without treating a single score as a scientific conclusion.
Hands-on example
Try this in FigureCheck
Fit the curved example, inspect residuals, and compare the fit-quality evidence with the visible curve.
Opens a small synthetic dataset with a step-by-step guide in a new editable session.Fit a selected series
- Open the Fit tab and select the visible series you want to model.
- Choose a model such as Linear, Polynomial, Exponential, Sine, or Fourier.
- For polynomial or Fourier models, set the degree or number of terms.
- Inspect the curve, residuals, validation signal, and scientific plausibility together.
Fits remain attached to their own series, so switching series does not combine separated groups or erase completed fits.
Read fit quality carefully
- R² describes how much variation the fitted curve explains in the selected data. It does not prove the model is correct.
- Pearson r describes linear association between X and Y. A curved relationship can be strong even when Pearson r is modest.
- Residuals show what the fit misses. Structured residual patterns often reveal a poor model.
- ML fit signal estimates underfitting or overfitting from validation behavior. Treat the probability and evidence as guidance, not a verdict.
Compare two series
In Stats, choose two visible series. FigureCheck runs distribution and effect-size checks only on those selected series. Use an independent comparison unless observations are intentionally matched pair by pair.
Statistical significance does not measure practical importance. Report the effect size, uncertainty, sample size, and study design alongside the p-value.