10.6 Dealing with Wrong Models

If Assumption 1 (Independence) may be faulty, we will later consider hierarchical Bayesian models in Unit 4 to handle dependent grouped data.

If Assumption 2 (Linearity) may be faulty, perhaps a transformation can help, such as a logarithmic scale for \(X\), \(Y\), or both.

If Assumption 3 (Normality) may be faulty, consider a different distribution (e.g. Poisson, binomial, negative binomial)