

   RReeggrreessssiioonn ffoorr aa ppaarraammeettrriicc ssuurrvviivvaall mmooddeell

        survreg(formula, data=sys.parent(), subset, na.action,
        link=c("log", "identity"),
        dist=c("extreme", "logistic", "gaussian", "exponential"),
        fixed, eps=0.0001, init, iter.max=10, model=F, x=F, y=F, ...)

   AArrgguummeennttss::

    formula: a formula expression as for other regression mod-
             els.  See the documentation for `lm' and `formula'
             for details.

       data: optional data frame in which to interpret the
             variables occuring in the formula.

     subset: subset of the observations to be used in the fit.

   na.action: function to be used to handle any NAs in the
             data.

       link: transformation to be used on the y variable.

       dist: assumed distribution for the transformed y vari-
             able.

      fixed: a list of fixed parameters, most often just the
             scale.  (When I implement the t-dist, it will
             include the degrees of freedom).

        eps: convergence criteria for the computation.  Itera-
             tion continues until the relative change in log
             likelihood is less than eps.

       init: optional vector of initial values for the
             paramters.

   iter.max: maximum number of iterations to be performed.

      model: if TRUE, the model frame is returned.

          x: if TRUE, then the X matrix is returned.

          y: if TRUE, then the y vector (or survival times) is
             returned.

        ...: all the optional arguments to lm, including `sin-
             gular.ok'.

   VVaalluuee::

        an object of class `survreg' is returned, which inher-
        its from class `glm'.

   CCoommppuuttaattiioonn::

        This routine is not as robust against nearly singular X
        matrices as lm(); the problem occurs when we explicitly
        invert the covariance matrix with solve().  This can
        sometimes be solved by subtracting the mean from all
        continuous covariates.

   EExxaammpplleess::

        survreg(Surv(futime, fustat) ~ ecog.ps + rx, fleming, dist='extreme',
                  link='log', fixed=list(scale=1))   #Fit an exponential

