why (not) linear regression?

why?

• simple
• easy to interpret (more important than you can imagine, especially when you need to explain the model to someone)

why not?

• relationship is never linear (well, almost never)
• example, usually wage varies linearly with age, but it kinda flattens up to a certain age
• try smooth spline, which would capture non-linearity.  
• Limit the degree of freedom to avoid excessive model variance (ie. over fitting ) 
• smooth.spline(age,wage,df=16) 
• "we can use LOO cross-validation to select the smoothing parameter for us automatically" 
• smooth.spline(age,wage,cv=TRUE)