jclbrooks / MD_Stream_Salamanders

MIT License
0 stars 0 forks source link

Little data in any year, difficult for dynamic model #14

Open djhocking opened 5 years ago

djhocking commented 5 years ago

######## Model is having lots of trouble using different values of psi by transect (site) for year one. There is very little data in that year ######## Options:

1. Use a single value of psi because all data that year from a single site - problem is the estimates from other sites will not be realistic or based on their information

2. Consider some way to combine information about b0 and b1 and relate it to psi so it comes from a common distribution or with some covariance structure

3. Rather than using calendar year, have psi start with the first year at any site - maybe could add calendar year as some sort of covariate

---- 3 seems like the best option but is certainly more problematic to implement - ugh :(

check the number of sites with observed salamanders by year

dfus_obs <- apply(dfus_3d, MARGIN = c(1, 3), max, na.rm = TRUE) dfus_obs[dfus_obs == -Inf] <- 0 cbind(sort(years), colSums(dfus_obs)) # no salamander observations in the first 2 years?!? Most years only a couple observations except 2012, 2015, and 2017 - something seems wrong with this but maybe only because DFUS. Other species better?

dfus_obs

how many surveys

sites_yr <- apply(dfus_3d, MARGIN = c(1, 3), max, na.rm = TRUE) sites_yr[sites_yr != -Inf] <- 1 sites_yr[sites_yr == -Inf] <- 0

sites_yr <- colSums(sites_yr) cbind(sort(years), sites_yr)

for trial, could throw out first 5 years and start with 2006 where 52 transects were surveyed