[R-sig-ME] Multiple random slopes in coxme

Mon Dec 7 01:38:37 CET 2020

Huh. I'll think about how that generalizes to the bigger use case. Thanks!

On Sun, Dec 6, 2020 at 6:50 PM David Duffy <David.Duffy using qimrberghofer.edu.au>
wrote:

> > I have placed a masked sample of my data as a .csv at
> > https://github.com/psmalone/reprex/blob/main/coxme_test.csv .
>
> I don't know how much it helps, but I think parametric survival models are
> more robust if you want slopes
> eg for your data above, a Weibull model:
>
> survreg(formula = Surv(FailTime, Event) ~ x1 + x2 + (x1 + x2) *
> frailty(cluster), data = x)
>
>                     coef     se(coef) se2      Chisq    DF   p
> (Intercept)          7.47245 0.038162 0.030435 38340.62  1.0 0.0e+00
> x1                  -0.04780 0.047612 0.044369     1.01  1.0 3.2e-01
> x2                   0.06143 0.038344 0.038165     2.57  1.0 1.1e-01
> frailty(cluster)                                 157.43 71.1 1.8e-08
> x1:frailty(cluster)  0.00104 0.000746 0.000659     1.95  1.0 1.6e-01
> x2:frailty(cluster) -0.00171 0.000642 0.000638     7.08  1.0 7.8e-03
>
> Scale= 0.552
>
> Iterations: 10 outer, 30 Newton-Raphson
>      Variance of random effect= 0.0522   I-likelihood = -5351.6
> Degrees of freedom for terms=  0.6  0.9  1.0 71.1  0.8  1.0  1.0
> Likelihood ratio test=224  on 74.3 df, p=<2e-16  n= 5000
>
>
>

-- 
Patrick S. Malone, Ph.D., Malone Quantitative
NEW Service Models: http://malonequantitative.com

He/Him/His

	[[alternative HTML version deleted]]