Hi,
I'm trying to using pspline in bic.surv{BMA}.
#############################
library(BMA)
library(survival)
data(veteran)
test.bic.surv<- bic.surv(Surv(time,status) ~
karno+pspline(age,df=3)+diagtime+prior, data = veteran, factor.type = TRUE)
summary(test.bic.surv, conditional=FALSE, digits=2)
#############################
The results are:
---------------------------------------------
Call:
bic.surv.formula(f = Surv(time, status) ~ karno + pspline(age, df = 3) +
diagtime + prior, data = veteran, factor.type = TRUE)
24 models were selected
Best 5 models (cumulative posterior probability = 0.59 ):
p!=0 EV SD model 1 model 2 model 3
model 4 model 5
karno 100.0 -3.3e-02 0.0051 -0.033 -0.032 -0.034
-0.033 -0.033
pspline.age..df...3.1 9.3 9.3e-02 0.3678 . . 1.068
. .
pspline.age..df...3.2 5.0 1.6e-02 0.1280 . . .
. 0.368
pspline.age..df...3.3 4.3 -1.6e-04 0.1126 . . .
. .
pspline.age..df...3.4 37.4 -3.5e-01 0.5445 . -0.917 .
. .
pspline.age..df...3.5 4.3 -5.9e-03 0.1028 . . .
. .
pspline.age..df...3.6 5.2 9.1e-03 0.0968 . . .
0.287 .
pspline.age..df...3.7 4.7 -7.5e-03 0.0983 . . .
. .
pspline.age..df...3.8 4.4 2.0e-02 0.2562 . . .
. .
pspline.age..df...3.9 4.4 -1.9e-02 0.2490 . . .
. .
pspline.age..df...3.10 4.4 -8.2e-02 1.0178 . . .
. .
diagtime 4.1 5.5e-05 0.0017 . . .
. .
prior 4.2 -3.1e-05 0.0043 . . .
. .
nVar 1 2 2
2 2
BIC -36.774 -35.832 -33.814
-32.651 -32.498
post prob 0.281 0.176 0.064
0.036 0.033
--------------------------------------------------
However, each column of basis matrix of age is treated as individual covariate
and BMA doesn't answer the question what's the probability of age being
included in the model.
I wonder if the bic.surv could provide the overall posterior probability for
each spline variable (similar to factor variables whose overall PMP is given).
Thanks you very much for your help!
Hui Xiong
[[alternative HTML version deleted]]
