The current process we use to fine-tune potential models relies on the
principle of least-squares (as opposed to the principle of maximum likelihood).
The principle of least-squares states that our goal is to find a model
that minimizes the "sum of squared errors." The sum of squared errors
is a measure of how well the values predicted by the model match
the actual data points. To calculate a model's sum of squared errors:
- Take the difference between each value predicted by the model
and the actual value. In the realm of correlation and regression,
the difference between
a predicted and actual value is sometimes referred to as a residual.
- Some of these values may be positive and some may be negative. Each
value is squared, and the result is known as a squared error or squared
deviation.
- When the squared error for all data points is summed, the result is a
single number that is referred to as a "sum of squared deviations," "sum of squared
errors," or sometimes a "sum of squared residuals."
The primary goal of the fine-tuning process is to
minimize the sum of squared errors while
keeping the number of predictor variables to a minimum.
So, when we are "fine-tuning" a model, we are
really focusing on finding the best coefficients for
each variable in a potential model, using the least-squares
principle to find coefficients for each variable in a model.
Take me back to the home page - I already know this
"stuff". :)
Generate your own models using our online tool, if you haven't already done so.
Is my online regression analysis done yet?