Linear models, as with any kind of models, require that we check their assumptions to justify their application. The accuracy and interpretability of the results comes from adhering to a model's assumptions. Sometimes these will be rigorous assumptions in the sense that if they are not strictly met, then the model is not considered to be valid at all. Other times, we will be working with more flexible assumptions in which a degree of criteria from the analyst will come into play.
For those of you interested, a great article about models' assumptions is David Robinson's, K-means clustering is not free lunch, 2015 (http://varianceexplained.org/r/kmeans-free-lunch/).
For linear models, the following are some of the core assumptions:
- Linearity: There is a linear relation among the variables
- Normality: Residuals are normally distributed
- Homoscedasticity...
