Nonlinear statistical modeling: On selecting regression variables to maximize their significance