[R] linear predictors and survreg function

[carol white]

Hi,
I calculated the linear predictors derived from weibull model using ovarian data sets. I calculated the linear predictors as the sum of covariates weighted by the weibull coefficients and compared to the linear predictors generated by survreg function. Why are they different? note that the first element of coefficients vector is intercept was excluded in my calculation.

Look forward to your reply,

Carol

--------------------------------------------------
data(ovarian)
library(survival)
survreg.obj = survreg(Surv(ovarian[,1],ovarian[,2])~age +resid.ds +rx +ecog.ps,ovarian, dist = "weibull", scale = 1)
> survreg.obj$linear.predictors
 [1] 5.298074 5.108976 5.558852 7.584172 7.221841 7.202655 7.019320 6.764081
 [9] 6.011550 7.939097 7.174129 8.634805 6.783737 7.261585 8.955989 8.366687
[17] 7.970807 8.489844 8.302639 8.385361 7.553247 4.855690 7.851908 7.235689
[25] 6.616655 7.917497

*******************

lp = survreg.obj$coefficients[2:5]%*%t(ovarian[,3:6])
> lp
             1         2         3         4         5         6         7
[1,] -7.484549 -7.673647 -7.223771 -5.198451 -5.560782 -5.579968 -5.763303
             8         9        10        11        12        13        14
[1,] -6.018542 -6.771073 -4.843526 -5.608495 -4.147818 -5.998886 -5.521038
            15        16        17        18        19        20        21
[1,] -3.826634 -4.415936 -4.811816 -4.292779 -4.479984 -4.397262 -5.229376
            22        23        24        25        26
[1,] -7.926933 -4.930715 -5.546934 -6.165968 -4.865126