
296 ◾ Software Metrics
cannot even explain the results without recourse to factors that do not appear
in the database. is is a key point: with causal models we seek to dig deeper
behind and underneath the data to explore richer relationships missing from
over-simplistic statistical models. In doing so, we gain insights into how best
to control risk and uncertainty. e regression model, based on the idea that
we can predict automobile crash fatalities based on temperature, fails to
answer the substantial question: how can we control or inuence behavior to
reduce fatalities. is at least is achievable; control of weather is not.
Statistica ...