Safety boundary
Use supplied datasets and approved simulations. Any physical demonstration must be teacher-supervised and explicitly classroom-safe; no improvised reactive mixtures.
Explicit chemistry instruction
A model should be judged by how well it predicts data, not by whether a curve looks familiar.
Residuals are observed minus predicted values. Random small residuals support adequacy better than patterned residuals.
Transformations can make relationships easier to compare, but the raw data must remain preserved.
Vocabulary: model fit; residual; transformation; prediction; evidence.
Worked example
Two models have similar average error, but one leaves a curved residual pattern while the other leaves random scatter; the second is usually better supported.
Guided question
Why can residual pattern matter more than a single fit score?
Independent work
- Fit two supplied simple models.
- Calculate residuals.
- Graph residuals.
- Rank the models and justify the choice.
Misconception check
A high correlation does not guarantee a correct causal or mechanistic model.
Formative check
What residual pattern suggests systematic model failure?
Reveal response
A nonrandom pattern such as curvature or consistent sign bias.
Mentor evidence conference
Ask Archaemenes to audit controls, competing predictions, residuals, uncertainty, and whether the conclusion exceeds the evidence.
Teacher guidance
Foundation: Use supplied prediction tables.
Core: Compare model residuals.
Extended: Discuss overfitting and alternative models.