# Library
library(R2jags)

# Read in data
app <- read.csv(paste0(getwd(), "/appendix.csv"), header = T)

# Write the model in BUGS Language as a function
lm1_jags = function() {
		
		# Likelihood
		for(i in 1:n) {
			betas[i] ~ dnorm(mu[i], tau.obs[i])
			tau.obs[i] <- pow(se[i], -2)
			mu[i] ~ dnorm(mu.int[fg[i]], tau.species)
		}
	 	
	 	# Priors 
	 		# - Beta estimates across each FG
		 	for(f in 1:7) {
		    	mu.int[f] ~ dnorm(0, 0.0001)
		 	}

	 		# - Error associated with beta estimates across each FG 
	 		sd.species ~ dunif(0, 100)
	 		tau.species <- pow(sd.species, -2)
}

# Specify inital values
init_values <- function() {
	list(mu.int = rnorm(7,1))
}

# Names of parameters to save
params = "mu.int"

# Bundle the data
	# Agricultural data (beta 4)
	test_data_agric <- list(
		'betas' = app$Paagric,
		'se' = app$Paagric_SE,
		'fg' = as.numeric(app$FG),
		'n' = dim(app)[1]
	)

	# Urban data
	test_data_urban <- list(
		'betas' = app$Paurban,
		'se' = app$Paurban_SE,
		'fg' = as.numeric(app$FG),
		'n' = dim(app)[1]
	)

# Fit the models
	# Agricultural land use meta-analysis
	fit_lm1_agric <- jags(data = test_data_agric, 
					 parameters.to.save = params,
					 model.file = lm1_jags,
					 n.chains = 3,
					 n.iter = 50000,
					 n.burnin = 25000,
					 n.thin = 10,
					 DIC = F
	)

	# Urban land use meta-analysis
	fit_lm1_urban <- jags(data = test_data_urban, 
					parameters.to.save = params,
					model.file = lm1_jags,
					n.chains = 3,
					n.iter = 50000,
					n.burnin = 25000,
					n.thin = 10,
					DIC = F
	)