Table 5B. Coefficients for the Model Predicting Colorectal Cancer in Women
VariableCoefficientStandard Error
Age*
    Linear component0.09000.0112
    Nonlinear component−0.02760.0114
Diabetes0.23330.0821
Regular use of multivitamins−0.16650.0568
Family history of colon cancer0.31590.0820
Years of education*
    Linear component0.07440.0204
    Nonlinear component−0.07570.0212
Race/ethnicity
    Hawaiian−0.24100.1284
    Japanese0.01400.0844
    Latino−0.39670.0977
    White−0.34060.0892
Body mass index
    Linear component0.00750.0146
    Nonlinear component0.01210.0177
Alcoholic drinks per day
    Linear component−0.08860.3046
    Nonlinear component0.40920.7886
Regular use of NSAIDs
    Yes, not currently−0.04640.0731
    Yes−0.23700.0887
Pack-years of smoking†
    Linear component0.06270.0498
    Nonlinear component 1−1.85152.1240
    Nonlinear component 22.29992.6795
Estrogen use
    Yes, not currently−0.04430.0719
    Yes−0.24500.0724

  • ↵* Continuous variables were modeled using restricted cubic splines with three knots. Therefore, continuous variables contain one coefficient for the linear component (displayed first) and a non-linear component (displayed in the next row and denoted with a single quotation mark).

  • ↵† Unable to obtain a natural spline using 3 knots and therefore restricted cubic splines with 4 knots were used for this variable.

  • NSAIDS, nonsteroidal anti-inflammatory drugs.