Machine Learning With Go

Linear regression assumptions and pitfalls

Like all machine learning models, linear regression does not work in all situations and it does make certain assumptions about your data and the relationships in your data. The assumptions of linear regression are as follows:

• Linear relationship: This might seem obvious, but linear regression assumes that your dependent variable depends on your independent variable linearly (by means of the equation for a line). If this relationship is not linear, linear regression will likely perform poorly.
• Normality: This assumption means that your variables should be distributed according to a normal distribution (which looks like a bell shape). We will come back to this property later in the chapter and discuss ...