Flexible Modeling

As other statistical modeling packages, Epicure supports logistic regression, matched case-control studies, Cox regression, and Poisson regression.

Epicure is designed to easily allow more flexible models than the often-used, but limited, loglinear models. Out of the box, Epicure supports:

Additive Models

T₀ + T₁ + T₂ + …

This allows absolute rates for each of the terms to be modeled.

Multiplicative Models

T₀ * (1 + T₁ )(1 + T₂ ) …

The “Excess Relative Risk” model (ERR). This allows linear increases proportional to the background rates (T₀) for each of the exposures (T_1,2,…). Attributable fractions for each of the exposure terms can be easily calculated.

Geometric Mixture Models

T₀ * (∏ T_1…)^θ (1 + Σ(T_1… – 1)^1-θ

The geometric mixture model is a combination of a multiplicative model (if _^θ= 1) or a product additive model (if _^θ= 0). The value of _^θ can be estimated to determine the nature of the response.

Future Flexibility

The 64-bit version of Epicure currently under development has been expanded to include these and other model forms:

T₀ (1 + ERR_rad + ERR_smk + θ ERR_rad ERR_smk)

Where θ is an interaction term.

And:

T₀ (1 + ERR_smk) (1+ ERR_int + ERR_ext)

where the smoking and radiation effects are multiplicative but the internal and external exposure effects are additive.

Hirosoft