Occupancy models in R

posts related to the RMark library, which may not be of general interest to users of 'classic' MARK

Occupancy models in R

Postby dhewitt » Thu Nov 08, 2007 5:24 pm

I know that Jeff Laake has provided a package RMark that links R and MARK. That's excellent, and if I'm reading the stars correctly, that will be the future. I am just wondering if anyone has tried fitting occupancy models using RMark yet, and if any issues have arisen.

I'm fairly well-versed in R, but haven't yet tried out RMark.

Thanks for any comments.

Evan, if you could toss this over to the MARK list as well that would be great.

Dave Hewitt
dhewitt
 
Posts: 150
Joined: Tue Nov 06, 2007 12:35 pm
Location: Fairhope, AL 36532

Not all MARK models are in RMark

Postby jlaake » Sat Nov 10, 2007 4:00 pm

RMark was originally designed for my own use with large CJS models and then I expanded it to recapure/recovery models again for my own use. Since I posted it for others to use, I have added some of the other models that folks have requested like closed models, JS (Pradel models), robust design, multi-state etc. However, RMark will not work with every model in MARK. I have written RMark in a way that allows models to be added fairly easily. As long as they aren't too different it is easy to add new ones but it still takes time to code, test and create an example.

I've not needed some of the non c-r type models but will add them if there is enough interest. I have conversed with Dave off-list and discussed creating an example for occupancy later next year. The following is the list of MARK models that are currently supported by RMark.

CJS Recaptures only
Recovery Recoveries only
Burnham Both(Burnham)
Barker Both(Barker)
Pradel Pradel recruitment only
Pradsen Pradel survival and seniority
Pradlambda Pradel survival and lambda
Pradrec Pradel survival and recruitment
LinkBarker Available only in change data type as Link-Barker
Closed Closed - no heterogeneity
HetClosed Closed – heterogeneity
FullHet Closed – full heterogeneity
Huggins Huggins – no heterogeneity
HugHet Huggins - heterogeneity
HugFullHet Huggins – full heterogeneity
POPAN POPAN
Known Known - known fate data (e.g, radio-tracking)
Multistrata Multistrata - CJS model with strata
Robust Robust design with Closed models for secondary periods - no heterogeneity
RDHet Robust design with Closed models for secondary periods – with heterogeneity
RDFHet Robust design with Closed models for secondary periods– full heterogeneity
RDHuggins Robust design with Huggins models for secondary periods – no heterogeneity
RDHHet Robust design with Huggins models for secondary periods - heterogeneity
RDHFHet Robust design with Huggins models for secondary periods – full heterogeneity
Nest Survival model
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Not all MARK models are in RMark

Postby cooch » Sat Nov 10, 2007 6:05 pm

jlaake wrote:RMark was originally designed for my own use with large CJS models and then I expanded it to recapure/recovery models again for my own use....


It is also worth noting that RMark is a sophisticated replacement interface for MARK. In using RMark, you use a different (not necessarily simpler) convention for building models (often a plus), you lose some of the familiar and convenient features of the the standard MARK gui (the results browser, for example), but underneath it all, the numerical routines in MARK.EXE do *all* the heavy lifting. RMark is just a different vehicle to pass models to, and retrieve results from, MARK. RMark does *not* do the numerical estimations.
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: Occupancy models in R

Postby cooch » Sat Nov 10, 2007 6:16 pm

dhewitt wrote:I know that Jeff Laake has provided a package RMark that links R and MARK. That's excellent, and if I'm reading the stars correctly, that will be the future.


Probably not. It would be 'the future' if everyone used R (in my Department, <5% of folks I can think of do...), and/or if it were 'generally' easier to use RMark as the interface to MARK (it is easier in some specific applications, but not at all for the MARK user at the beginning or intermediate level, since you're often simply substituting the visual complexity of building a DM graphically in MARK with the often obtuse and arcane conventions of the R language - it is not hard to come up with examples of RMark code - or R in general - which are virtually impenetrable to someone without a significant R background).

I can teach someone to use the 'classic' MARK interface even if they have no background in R (or any other statistical programming environment). Further, using the classical interface - especially the graphical DM - forces the user to actually think about what they're doing (you learn a heck of a lot more about linear models by actually confronting the DM directly, rather than through some 'canned' function). For highly experienced analysts (like Jeff, obviously), this isn't useful, and is in fact annoying - RMark lets you dispense with building the DM manually.

But, to learn to use RMark effectively, you need to learn to be fairly proficient in R. Perhaps that is value added, but it is an extra step.

RMark is outstanding - I use it fairly often. It is an extremely useful addition to the 'arsenal' for working with data from marked individuals.
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University


Return to RMark

Who is online

Users browsing this forum: Google [Bot] and 13 guests

cron