Design for multi-state model with live and dead recoveries

[hanchinp]

I am new to MARK, but searched the forums and the gentle introduction (did not read in its entirety) and could not asure myself of how to properly set up this capture history file.

Our goal is a multi-state model with 10 locations and over 10,000 individuals (walleye) in which we want to understand movement and obtain estimates of abundance (relative). We marked fish on various occasions over 2 years, though most took place in the spring of 2 years. We have both live and dead recoveries that have occurred during surveys throughout the year and from anglers. Also, some anglers provide a recovery for a fish that was not harvested.

I initially thought that defining 3-month capture periods would be best since shorter periods would likely result in very low probabilites. One problem encountered is that numerous fish are captured multiple times in the same period (e.g. twice alive, or once alive, once dead). Can I simply use the dead recovery when a fish is encountered alive in the same period (e.g ), or should I use a design that allows for the coding of both captures (alive and dead)? Since probabilites are estimated between periods, does it add anything to have the live capture coded for a given period (e.g. LDLDLDLD)? If I define the spawning period as a shorter period, it will eliminate many of these "problem" capture histories, but some will remain.

I am still at the stage of simply getting the data from Access to an apprpriate capture history file and am looking for any suggestions.

[claudiapenaloza]

I'm not sure I understood how frequent your sampling was... but if you have repeated live encounters of individuals within the period of time you call a "sampling period", you could use a Robust Design to analyze your data. You would subdivide your primary sampling period into secondary sessions during which you have repeated live encounters of your individuals. You would still assume death occurs between your live primary sampling periods (read the Gentle Introduction Chapter on Robust Design to become more familiar with it). In MARK you would use a “Multistate Live and Dead enc.” and coerce it to a robust design or a “Closed Robust Design Multistate” with your 10 live states and an additional dead state and then coerce it to separate transition from survival.
It may seem a little complicated, and your system may not warrant the extra effort, but if you think there is any chance of temporary emigration (not all your individuals are present in the sampling area at a given time) you should use a robust design.
Otherwise, your initial suggestion of shortening your primary sampling periods such that each encounter falls in a different period is a much simpler solution…
And remember, the more captures the better! (i.e., an “extra” live encounter is never bad).