# Solved – How to get convergence using coxph (R) given that model converges using proc phreg (SAS)

I have a data set

``> head(data)   id centre        u      time event x c1 c2 c3 c4 c5 c6 c7 c8 c9 c10 c11 c12 c13 c14    1      1 0.729891 0.3300478     1 0  1  0  0  0  0  0  0  0  0   0   0   0   0   0    2      1 0.729891 7.0100000     0 0  1  0  0  0  0  0  0  0  0   0   0   0   0   0    3      1 0.729891 7.0150000     0 1  1  0  0  0  0  0  0  0  0   0   0   0   0   0    4      1 0.729891 1.3616940     1 0  1  0  0  0  0  0  0  0  0   0   0   0   0   0    5      1 0.729891 7.0250000     0 0  1  0  0  0  0  0  0  0  0   0   0   0   0   0    6      1 0.729891 5.0824055     1 0  1  0  0  0  0  0  0  0  0   0   0   0   0   0 ``

and I want to fit a proportional hazards model.

IN R

I first defined `formula`:

``> formula Surv(time, event) ~ x + c1 + c2 + c3 + c4 + c5 + c6 + c7 + c8 +      c9 + c10 + c11 + c12 + c13 + c14 ``

and I used the `coxph` function:

``> library(survival) > mod <- coxph(formula, data=data) Erreur dans fitter(X, Y, strats, offset, init, control, weights = weights,  :    NA/NaN/Inf dans un appel à une fonction externe (argument 6) De plus : Message d'avis : In fitter(X, Y, strats, offset, init, control, weights = weights,  :   Ran out of iterations and did not converge ``

IN SAS

I used `proc phreg`:

``proc phreg data=data;     model time*event(0) = x c1-c14 / ties=breslow; run; ``

Partial output:

``                      Convergence Status                       Convergence criterion (GCONV=1E-8) satisfied. ``
Contents

### Question:

• Do you have any suggestion to make it work in R?

I have tried to increase `iter.max` but the problem is still there… If needed, I can provide the data (600 rows).

What about changing of the starting values? Supply to `init` similar values to the output of SAS and see what happens. This in general is not a good strategy to get the convergence, but it helps to see whether the problem is in algorithm, or just in starting values.