Page 1 of 1

design of pi matrix in the mvmscjs model

PostPosted: Mon Feb 26, 2018 1:56 pm
by JaumeForcada
Hello, I am trying to run a model, very similar to the sealions data set example analysis in marked. In that analysis there are 8 strata (A++, A+-, A-+, A–, S++, S+-, S-+, S–), from site A and S, and right tag status + or - and left tag status + or -, which can be unknwon (u). By default, I imagine that the design function assumes that as a convention, the fully known observation (e.g., A++) is specified as reference cell for delta. There is no specification for pi, because all seals are expected to start in site A and tag status ++. How can we define pi if that is not the case for every seal at capture? Some seals may start in A+- (or A-+) and others at N+- (or N-+). Is there a simple way to do that? Many thanks in advance

Re: design of pi matrix in the mvmscjs model

PostPosted: Mon Feb 26, 2018 4:08 pm
by dsjohnson
Since the initial debut of the 'mvmscjs' model in marked, Jeff has added a pi modeling capabilities. In the design data list there is a 'pi' design data matrix. The fix column in the ddl$pi matrix will have a 1 for the first state observed for each animal and 0s everywhere else. So, as long as you fully observe the state on initial marking, you should be fine as is. After you create the design data with 'process.design.data()' you can always check the pi$fix column to make sure. However, if the state is not fully known at the time of first marking, then you can place NAs in the fixed column for states that are possibilities on initial marking.

Re: design of pi matrix in the mvmscjs model

PostPosted: Tue Feb 27, 2018 5:27 am
by JaumeForcada
This is great, many thanks for your help, and Jeff's also for making life easier!