ERROR -- Multiple strata within one primary session

[j.harv3y]

Hi, I'm running an MSORD, with weekly sampling intervals as secondary sessions, and years as primary sessions. My states are 1 (encountered) and U (unobserved).

When I try to run the models I get the following error;
ERROR -- Multiple strata within one primary session.
STOP ERROR

I haven't had this before (when I have run similar models which an individual has been U in one secondary and then 1 in another secondary within the same primary sessions it's been fine.I don't understand the error and can't seem to work out how to fix this. I'm incredibly stuck!

Any advice would be really appreciated.

Thanks so much,

Jess

[Bill Kendall]

You're saying that you have an individual that is in two states within a primary period? You can't have that. State membership is assumed to be static within a primary period. I'm also not sure why you would have a 'U' in your history if this is an unobservable state (i.e., p(U)=0).

If before you were using the MSORD model with state uncertainty, then a lowercase 'u' would denote the individual was observed but the state is unknown. That's a different situation.

[j.harv3y]

You're saying that you have an individual that is in two states within a primary period? You can't have that. State membership is assumed to be static within a primary period. I'm also not sure why you would have a 'U' in your history if this is an unobservable state (i.e., p(U)=0).

If before you were using the MSORD model with state uncertainty, then a lowercase 'u' would denote the individual was observed but the state is unknown. That's a different situation.

Hi Bill,

Thanks a million for your reply. From reading the mark handbook I was under the impression from 15.8.2 that 'u' denotes the individual was not seen and therefore the state is unknown? So you could use 0/'u' interchangeably.

Do you know is there an option to use just an ORD rather than MSORD in rmark? From what I've read these seem to be used interchangeably in the book.


Really appreciate your thoughts on this,

Jess

[Bill Kendall]

So, if you have a case where you do have more than one state of interest that is observable, and you have cases where the individual is detected but its state is unknown (designated by the 'u'), then you would use the model in Multi-state with State Uncertainty -> 1 uncertain state open (robust design) and go from there (it would include pi and delta parameters for accounting for the state uncertainty).

If you just have an extra state that is unobservable, or you really just have one state but you need an extra state because the model requires at least 2, then your data should just be 1's (or whatever you want to use) and 0's.

[j.harv3y]

So, if you have a case where you do have more than one state of interest that is observable, and you have cases where the individual is detected but its state is unknown (designated by the 'u'), then you would use the model in Multi-state with State Uncertainty -> 1 uncertain state open (robust design) and go from there (it would include pi and delta parameters for accounting for the state uncertainty).

If you just have an extra state that is unobservable, or you really just have one state but you need an extra state because the model requires at least 2, then your data should just be 1's (or whatever you want to use) and 0's.

Bill, You're incredible!!

Can I just clarify one last thing please? So I will do as you suggested with just having the 1's and 0's, as my state is just 'encountered' essentially. The model as you said requires two states, so if I tell it the states are '1' and '0', will Rmark know that 0 is 'dead/ not encountered' or by assigning 0 to a state will the 0 for not encountered be overruled and then will rmark think that the individual has been encountered in a specific state (as above with using 'u')?

Thank you so much again, really appreciate the help