Bug in transition probability matrix

[calsmallory]

Hello,

I'm tring to fit a multi-state model and I specified the transition probabilities in the Q matrix. In my dataset the only possible transitions are
1 -> 2
2 -> 3
2 -> 4
3 -> 2
3-> 4
however when the multistate model has converged the resulting pmatrix of the transitions probabilities shows values that are inconsistent with the allowed propabilities 1 -> 4 is 0.3235 when it should be 0.

kindly, can you look into this bug ? :)

With many thanks

[chjackson]

This is not a bug, but a common confusion between continuous-time and discrete-time models.

msm works in continuous time, and you have allowed instantaneous transitions in continuous time which allow a person to travel on pathways 1 -> 2 <-> 3 -> 4 or 1 -> 2 -> 4

The transition probability matrix is the probability that someone is in state s t years later, given they are in state r now.

So in your model, it is possible for an individual to be in state 1 at time 0, then in state 4 at time 30. They must have gone through states 2 and/or 3 at some time between 0 and 30.

[calsmallory]

Thank you very much for your quick reply is it clear now.