ttt.df<-data.frame(time=ttt,status=ttt.status)
 parameters(ttt.df)<-list(alpha=.3,gam.par=1)

 # - Weibull neg.-log-likelh with gradient and hessian I
 weib.nllh2<-function(status,t.time,
    alpha,gam.par) {
    DD<-sum(status)
    t.time.dead<-t.time[status==1]

    value<- -DD*log(alpha) -DD*log(gam.par) -
             (gam.par-1)*sum(log(t.time.dead)) +
             alpha*sum(t.time^gam.par)

    attr(value,"gradient") <- c(
      -DD/alpha + sum(t.time^gam.par),
      -DD/gam.par - sum(log(t.time.dead)) +
         alpha*sum(log(t.time)*t.time^gam.par))

    attr(value,"hessian") <- rbind(
      c(DD/(alpha^2),sum(log(t.time)*t.time^gam.par)),
      c(sum(log(t.time)*t.time^gam.par),
        DD/(gam.par^2) +
        alpha*sum( (log(t.time))^2 * t.time^gam.par )))
    value
 }
 w.fit2 <- ms(~weib.nllh2(status=ttt.df$status,
               t.time=ttt.df$time,alpha,gam.par),
               start=list(alpha=.1,gam.par=1.0),
               data=ttt.df)
 w.fit2
 value: 16.87063
 parameters:
       alpha  gam.par
 0.02871909 1.864433

 w.hess<-w.fit2$hessian
 w.hess
          [,1]     [,2]
 [1,] 7274.6137  0.00000
 [2,]  381.6464 22.67202

 w.hess[1,2]<-w.hess[2,1]  # set up diag. element since lower diag.
                           # returned for the hessian

 w.vcov<- solve(w.hess)    # estimated variance-covariance matrix

 se.alpha<- sqrt(w.vcov[1,1])
 se.gam.par<- sqrt(w.vcov[2,2])

 se.alpha
 [1] 0.03429527

 se.gam.par
 [1] 0.6143192