Need help with predictms for multistate modelling

[Andrew Cole]

Hi, please I need help with using the predictms command to derive predictions from a multi-state model, after using stmerlin. I keep getting the error "At least one prediction type must be specified". I have read the documentation and I am not quite sure what I am not doing correctly. I am using Stata 17. My codes and outputs are below:

 msset, id(id) states( incar ad ) times( incar_time ad_time ) covariates( age  gender )
variables age_trans1 to age_trans3 created
variables gender_trans1 to gender_trans3 created

mat tmat = r(transmatrix)

 mat list tmat

tmat[3,3]
               to:    to:    to:
            start  incar     ad
from:start      .      1      2
from:incar      .      .      3
   from:ad      .      .      .

stset _stop, enter(_start) failure(_status=1)

Survival-time data settings

         Failure event: _status==1
Observed time interval: (0, _stop]
     Enter on or after: time _start
     Exit on or before: failure

--------------------------------------------------------------------------
      3,818  total observations
          0  exclusions
--------------------------------------------------------------------------
      3,818  observations remaining, representing
      1,612  failures in single-record/single-failure data
 28,706.218  total analysis time at risk and under observation
                                                At risk from t =         0
                                     Earliest observed entry t =         0
                                          Last observed exit t =  25.10335

stmerlin age if _trans1==1, distribution(weibull) nolog

Fitting full model:

Survival model                                           Number of obs = 1,556
Log likelihood = -2211.3467
------------------------------------------------------------------------------
             | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
_t:          |            
         age |  -.3439441   .0202445   -16.99   0.000    -.3836226   -.3042656
       _cons |   3.423091   .2881495    11.88   0.000     2.858328    3.987854
  log(gamma) |  -.6763313    .033129   -20.42   0.000    -.7412628   -.6113997
------------------------------------------------------------------------------

estimates store m1

stmerlin age if _trans2==1, distribution(weibull) nolog

Fitting full model:

Survival model                                           Number of obs = 1,556
Log likelihood = -1618.1694
------------------------------------------------------------------------------
             | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
_t:          |            
         age |  -.0763549   .0322178    -2.37   0.018    -.1395007   -.0132092
       _cons |  -3.006277    .500387    -6.01   0.000    -3.987017   -2.025536
  log(gamma) |   .2422271   .0420599     5.76   0.000     .1597913     .324663
------------------------------------------------------------------------------

estimates store m2

stmerlin age if _trans3==1, distribution(weibull)nolog
note; a delayed entry model is being fitted

Fitting full model:

Survival model                                             Number of obs = 706
Log likelihood = -1703.1241
------------------------------------------------------------------------------
             | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
_t:          |            
         age |   -.105471   .0250724    -4.21   0.000     -.154612     -.05633
       _cons |  -1.532095   .3861284    -3.97   0.000    -2.288893   -.7752976
  log(gamma) |   .2845562   .0459795     6.19   0.000     .1944381    .3746743
------------------------------------------------------------------------------

estimates store m3

predictms, transmatrix(tmat) models(m1 m2 m3) at1(age 10)
At least one prediction type must be specified
r(198);

[Denise Vella]

Did you ever find an answer to this? Have the same issue

[Sharon Ayayo]

Hi, you need to specify the statistic you want, for example if you need transition probabilities then specify "probability" option, if you need transition-specific hazard function, then you need to specify "hazard" option. You need to specify at least one of the 7 options for the statistics that are available from the help file, you can also specify more than one statistic. See help predictms