**Why Maximizing R**^{2} Is Not the
Best Way To Pick A Non-Linear Equation
The data plotted at right was
simulated using an equation y = x^{2} with some added random errors.
**Clearly an exponential regression is the most
appropriate type.**
While **the 6th order polynomial has the best R2 value**, the
equation is unnecessarily complex and does not reflect reality as well as
the power regression line. |