Things2Implement.txt

workflow document:

?SCRbayes should give a quick "how to get started". I have some SCR
data and need an answer fast! HELP!!!!

SCRbayes exists to do Bayesian analysis of a certain class of SCR
models. (to be described shortly). But first, why should you use
SCRbayes over WinBUGS or JAGS or the R package secr (which does
likelihood analysis)? The main reason is that SCRbayes is relatively
more robust (in an operational sense) to very large problems. If you
have a lot of traps, sample occassions, or a large state-space grid,
SCRbayes is less likely to run out of memory or take forever to
produce estimates.  
Note: SCRbayes works strictly with a DISCRETE state-space grid. IF you
can define your state-space by a rectangle or a constrained rectangle
then JAGS or WinBUGS can be extremely fast. 
Note: secr can be very efficient too if you do not define the
state-space grid ("habitat mask" in secr terminology) or if you define
it to be relatively coarse. Its only when the state-space grid becomes
enormous that secr starts to break down [note: this is subject to
change]. 
Note: also when there are many, many traps (e.g., > 1000) both secr
and JAGS/WinBUGS start having problems. 

I have run models effectively in SCRbayes with 8000 traps, 8000
state-space grid points, and 144 sample occasions. These take 10-12
hours to run on a really fast computer but it works.

Note: multiple core capabilities will be developed. 
Note: use the "dumprate" argument (not yet implemented) to get
intermediate output. 
Note: how long should the MCMC algorithm run?


1. 
you need: trap locations in some coordinate system preferably UTM or other
Cartesian system. 

?traps
?TDF
need as.TDF
class "TDF"

2. 
you need: a state-space grid which defines the possible locations of
individuals. The grid should be a regular spaced grid of points. This
will be a nG x 2 matrix.  
You need to add a 3rd column which defines whether each point is
available habitat (1) or not (0). We recommend defining everything to
be habitat even if you think it isn't.

You should convert your statespace grid to an object of class
"statespace" using the as.statespace() function. 

If you don't have a state-space grid you can easily make one using the
make.statespace function. 

see ?statespace or ?make.statespace 

3. 
you need some encounter data from an actual study that you did.....
as.EDF


short term objectives:
 1. beef up scrData capabilities and print.scrdata

 2. move all data object handling from SCRh into scrData

 3. move GoF calculation into a helper function REMOVE from SCRh

 4. BEEF up print method for objects of class scrdata and scrfit


need print.statespace or print.ss
need to class statespace files. 
need as.statespace to coerce any nG x 2 matrix into a statespace file
need example in make.statespace helpfile
make.statespace needs to accept home range size as input 


Change function call from SCRf.fn to scrbayes() 



Class for "scrFitBayes" or "scrmcmc"? or....???

Goodness-of-fit stuff

Simulator function



Dead encounter data (location information) 



DONE      Pixel area (of state-space point) and monitor density.