Hello Chris,
I am trying to fit a 4 state multi-state model using hidden markov routine in
the msm package.
1. initial parameters:
twoway4.q <- rbind(c(0, 0.25, 0, 0.25), c(0.166,
0, 0.166, 0.166), c(0, 0.25, 0, 0.25), c(0,
0, 0, 0))
ematrix <- rbind(
c(0, 0.01, 0, 0),
c(0.01, 0, 0.01,0),
c(0, 0.1, 0, 0),
c(0, 0, 0, 0))
2. the model:
msm_covariates_sexandage <- msm(state ~ duration, subject = SerialNo,
data = Data, qmatrix = twoway4.q, ematrix = ematrix,death = 4, obstrue =
firstobs, covariates = ~ sex + age, method = "BFGS", control = list(reltol =
1e-10, fnscale = 50000, trace=1,REPORT=1))
msm stops after only 100 iterations and fails to compute the Hessian (See the
output below). Why is it stopping at 100 only iterations yet i have specified
reltol = 1e-10?
One additional information is that when i exclude age, the model converges
without any problem. I have tried using age groups instead of actual age but no
convergence can be achieved.
I have also tried subsetting the data in various biologically meaningful groups
but there would be no convergence.
3. the R output:
> msm_covariates_sex <- msm(state ~ duration, subject = SerialNo,
+ data = Data, qmatrix = twoway4.q, ematrix = ematrix,death = 4, obstrue =
firstobs, covariates = ~ sex + age, method = "BFGS", control = list(reltol =
1e-10, fnscale = 50000, trace=1,REPORT=1))
initial value 0.567920
iter 2 value 0.547379
iter 3 value 0.463524
iter 4 value 0.439987
iter 5 value 0.426417
iter 6 value 0.420621
iter 7 value 0.417181
iter 8 value 0.414153
iter 9 value 0.410340
iter 10 value 0.406045
iter 11 value 0.403618
iter 12 value 0.403025
iter 13 value 0.402781
iter 14 value 0.402672
iter 15 value 0.402653
iter 16 value 0.402630
iter 17 value 0.402560
iter 18 value 0.402429
iter 19 value 0.402221
iter 20 value 0.402051
iter 21 value 0.401978
iter 22 value 0.401944
iter 23 value 0.401903
iter 24 value 0.401811
iter 25 value 0.401622
iter 26 value 0.401311
iter 27 value 0.400978
iter 28 value 0.400814
iter 29 value 0.400788
iter 30 value 0.400780
iter 31 value 0.400759
iter 32 value 0.400714
iter 33 value 0.400602
iter 34 value 0.400372
iter 35 value 0.400056
iter 36 value 0.399865
iter 37 value 0.399761
iter 38 value 0.399743
iter 39 value 0.399706
iter 40 value 0.399636
iter 41 value 0.399464
iter 42 value 0.399158
iter 43 value 0.398792
iter 44 value 0.398584
iter 45 value 0.398524
iter 46 value 0.398516
iter 47 value 0.398490
iter 48 value 0.398478
iter 49 value 0.398443
iter 50 value 0.398400
iter 51 value 0.398308
iter 52 value 0.398223
iter 53 value 0.398222
iter 54 value 0.398192
iter 55 value 0.398190
iter 56 value 0.398164
iter 57 value 0.398163
iter 58 value 0.398162
iter 59 value 0.398159
iter 60 value 0.398151
iter 61 value 0.398136
iter 62 value 0.398115
iter 63 value 0.398099
iter 64 value 0.398093
iter 65 value 0.398092
iter 66 value 0.398091
iter 67 value 0.398089
iter 68 value 0.398082
iter 69 value 0.398066
iter 70 value 0.398033
iter 71 value 0.397984
iter 72 value 0.397950
iter 73 value 0.397936
iter 74 value 0.397934
iter 75 value 0.397933
iter 76 value 0.397927
iter 77 value 0.397919
iter 78 value 0.397906
iter 79 value 0.397897
iter 80 value 0.397894
iter 81 value 0.397894
iter 82 value 0.397893
iter 83 value 0.397891
iter 84 value 0.397887
iter 85 value 0.397877
iter 86 value 0.397855
iter 87 value 0.397818
iter 88 value 0.397772
iter 89 value 0.397731
iter 90 value 0.397715
iter 91 value 0.397710
iter 92 value 0.397706
iter 93 value 0.397701
iter 94 value 0.397697
iter 95 value 0.397689
iter 96 value 0.397668
iter 97 value 0.397645
iter 98 value 0.397628
iter 99 value 0.397611
iter 100 value 0.397597
final value 0.397597
stopped after 100 iterations
Warning message:
In msm(state ~ duration_in_years, subject = SerialNo, data = type1Data, :
Could not calculate asymptotic standard errors - Hessian is not positive
definite. Optimisation has probably not converged to the maximum likelihood
>
The University of Dundee is a registered Scottish Charity, No: SC015096
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
