Checking a model against evidence

Choosing between linear and exponential models by how the data grow is part of checking whether a model fits.

After solving a quadratic or exponential model, check its solution against the context and the data.

Residuals from a regression line are used to decide whether the model should be kept or revised.

Residuals show whether a model fits the data, and model validation uses them to decide whether to keep or change a model.

A trend line on a scatterplot is a model, and model validation checks how well it fits the data.

Residual patterns apply to any model, including a sine or cosine model whose midline, amplitude and period are checked against measurements.

A decision made by expected value is only as good as the probabilities in the model, which is why a model is checked against data.

An empirical distribution of results is compared with the model's theoretical distribution to check the model.

Asking whether a model's pattern could come from chance uses the same logic as a randomization test.

A model is checked by simulating it and comparing the simulated outcomes with the data.

Asking whether a report's evidence supports its claim is the same question as checking whether a model's predictions agree with the data.