####******************************************************************###
##Sample R code to accompany research article:                        ###
##Modeling to Predict Cases of Hantavirus Pulmonary Syndrome in Chile ###
##Monday December 30th, 2013                                          ###
####******************************************************************###

library(forecast)
library(TSA) 

#Fitting models with ARIMA errors in R

#Read in data
hanta <- read.table("/data path/hanta_data.txt")
climate <- read.table("/data path/climate_data.txt", header=TRUE)


#convert to time series variable
cases <- ts(hanta[,2], frequency=12, start=c(2001,1)) #second column of hanta contains case counts


#combine variables with one month lag
xresp <-  cbind(cases, c(NA, precip[1:143])) #HPS cases and precipitation

#fit and forecast in for loop
valls <- matrix(0, 36, 4)     #define matrix for later use
t = 1                         #counter 

for(months in 108:143){ 
        
        fitt2 <- Arima(xresp[,1], xreg=xresp[,2], order=c(1,0,1))  #fit ARIMA model

        j = months+1 

	pval <- forecast(fitt2, xreg=xresp[months:j,2], level=c(95))$mean   #predicted value
	plow <- forecast(fitt2, xreg=xresp[months:j,2], level=c(95))$lower  #lower CI value
	pupp <- forecast(fitt2, xreg=xresp[months:j,2], level=c(95))$upper  #upper CI value
	
	valls[t,1] <- pval
	valls[t,3] <- plow
	valls[t,4] <- pupp
	
	t = t+1; #update counter
	
}

valls