Normal and Lognormal Regression

In Normal and Lognormal Regression model, it is assumed that the survival times (or log survival times) originate from a normal distribution; the resulting model is basically identical to the ordinary multiple regression model, and can be defined as:

t = a + b1*z1 + b2*z2 + ... + bm*zm

where t denotes the survival times.

If log-normal regression is requested, t is replaced by its natural logarithm. The normal regression model is particularly useful because many data sets can be transformed to yield approximations of the normal distribution. Thus, this is the most general and fully parametric model (as opposed to Cox's proportional hazard model which is non-parametric), and estimates can be obtained for a variety of different underlying survival distributions.

Goodness-of-fit
The Chi-square value is computed as a function of the log-likelihood for the model with all independent variables (L1), and the log-likelihood of the model in which all independent variables are forced to 0 (zero, L0).