Hey @afredston, sending along a longer email shortly, but wanted to make sure to get this PR done. Hopefully everything makes sense from the new R script vast_fit_summer_flounder.R, which generates the biomass datasets needed to fit a VAST model that has spatial, and spatio-temporal variability turned on, and then a random walk autoregressive structure for the intercepts and the spatio-temporal variability component. The model is fit to data from 1972-2006 and then used to predict to 2006-2016 -- this is handled using the Pred_TF argument in the call to fit_model. The model run time isn't too horrible -- about 8 hours on my MacBook Pro with bias correction turned on. Along with the usual VAST stuff, I added in some parts to get stratified biomass indices for the different latitude bins/patches and then give a few example results plots based on some functions I have written. I think things look pretty good with the exception of the center of gravity part -- I haven't discovered yet what is causing it to peg things to the very south of the survey for most of the years, and it seems somewhat counter to what is seen when you look at the maps of predicted density. For these plots, you might just need to change to a shapefile for the land on your local computer as I didn't want to risk including those and getting the dreaded error about file size being too big for Git/GitHub. More in email soon, along with a copy of the fitted model object in case you want to skip right to looking at diagnostics/results :)
Hey @afredston, sending along a longer email shortly, but wanted to make sure to get this PR done. Hopefully everything makes sense from the new R script
vast_fit_summer_flounder.R, which generates the biomass datasets needed to fit a VAST model that has spatial, and spatio-temporal variability turned on, and then a random walk autoregressive structure for the intercepts and the spatio-temporal variability component. The model is fit to data from 1972-2006 and then used to predict to 2006-2016 -- this is handled using thePred_TFargument in the call tofit_model. The model run time isn't too horrible -- about 8 hours on my MacBook Pro with bias correction turned on. Along with the usual VAST stuff, I added in some parts to get stratified biomass indices for the different latitude bins/patches and then give a few example results plots based on some functions I have written. I think things look pretty good with the exception of the center of gravity part -- I haven't discovered yet what is causing it to peg things to the very south of the survey for most of the years, and it seems somewhat counter to what is seen when you look at the maps of predicted density. For these plots, you might just need to change to a shapefile for the land on your local computer as I didn't want to risk including those and getting the dreaded error about file size being too big for Git/GitHub. More in email soon, along with a copy of the fitted model object in case you want to skip right to looking at diagnostics/results :)