State Occupancy Probabilities for First-Order Markov Ordinal Model from a Model Fit

```
soprobMarkovOrdm(
object,
data,
times,
ylevels,
absorb = NULL,
tvarname = "time",
pvarname = "yprev",
gap = NULL
)
```

if `object`

was not a Bayesian model, a matrix with rows corresponding to times and columns corresponding to states, with values equal to exact state occupancy probabilities. If `object`

was created by `blrm`

, the result is a 3-dimensional array with the posterior draws as the first dimension.

- object
a fit object created by

`blrm`

,`lrm`

,`orm`

,`VGAM::vglm()`

, or`VGAM::vgam()`

- data
a single observation list or data frame with covariate settings, including the initial state for Y

- times
vector of measurement times

- ylevels
a vector of ordered levels of the outcome variable (numeric or character)

- absorb
vector of absorbing states, a subset of

`ylevels`

. The default is no absorbing states. (numeric, character, factor)- tvarname
name of time variable, defaulting to

`time`

- pvarname
name of previous state variable, defaulting to

`yprev`

- gap
name of time gap variable, defaults assuming that gap time is not in the model

Frank Harrell

Computes state occupancy probabilities for a single setting of baseline covariates. If the model fit was from `rms::blrm()`

, these probabilities are from all the posterior draws of the basic model parameters. Otherwise they are maximum likelihood point estimates.