## install the following packages
#install.packages("asnipe") 
#install.packages("Matrix") 
#install.packages("survival")
#install.packages("combinat") 
#install.packages("foreach")
#install.packages("doParallel")
#install.packages("matrixStats)


# set working directory that contains the file 'NBDA code 1.2.15.r' and the 'error_and_sensitivity_function parallel.r'
setwd("C:/Users/bssw/Desktop/Leeds University/PhD/Simulations - sensitivity of NBDA/Online Resources")
source("OR3_NBDA code 1.2.15.r") # sources functions
source("OR2_sensitivity functions.R") # sources functions 

# A)  read simulated data set (observation x individuals matrix)
sightings <- read.csv("OR5_simulated observational data.csv")

# B) extract how many times each dyad has been seen together and how many times they have been seen apart
 # these two matrices will be used to create a social network with observational error using a Bayesian approach.

together_apart_output <- together_apart(sightings=sightings)
write.csv(together_apart_output[,,1], file="together_matrix.csv") # save 'together' matrix
write.csv(together_apart_output[,,2], file="apart_matrix.csv") # save 'apart' matrix


# C) determine how many times individuals have been seen and choose cut-off points
cutoff <- sort(unique(colSums(sightings))) # extract colSums (=number of sightings)
cutoff <- cutoff[-length(cutoff)] # remove the maximum number of sightings (as there have to be at least 2 individuals that make the cut), otherwise association matrices cannot be calculated and the function returns an error. 
cutoff <- cutoff[-length(cutoff)] # remove second last as well

# D) run a first simulation with dropping learners and s=8 (social learning)

sim1 <- sensitivity_NBDA_ind_error(x=together_apart_output, # with s=8 and dropping learners
                                   sightings=sightings,
                                   cutoff=cutoff, 
                                   association_index="SRI", 
                                   iterations=10000, 
                                   s=8, 
                                   num_ind_learn=20, 
                                   cores=2,
                                   keep_learners = FALSE,
                                   delta_AICc = 2)

save(sim1, file="sensitivity_NBDA_with_error_false neg_drop learners.RData") # save the sim1 object 
write.csv(sim1$raw,"sensitivity_NBDA_with_error_false neg_drop learners_RAW.csv" )
write.csv(sim1$summary,"sensitivity_NBDA_with_error_false neg_drop learners_SUMMARY.csv" )

# E) run a second simulation with keeping learners and s=8 (social learning)
sim2 <- sensitivity_NBDA_ind_error(x=together_apart_output, # with s=8 and keeping learners
                                   sightings=sightings,
                                   cutoff=cutoff, 
                                   association_index="SRI", 
                                   iterations=10000, 
                                   s=8, 
                                   num_ind_learn=20, 
                                   cores=2,
                                   keep_learners = TRUE,
                                   delta_AICc=2)

save(sim2, file="sensitivity_NBDA_with_error_false neg_keep learners.RData") # save the sim2 object 
write.csv(sim2$raw,"sensitivity_NBDA_with_error_false neg_keep learners_RAW.csv" )
write.csv(sim2$summary,"sensitivity_NBDA_with_error_false neg_keep learners_SUMMARY.csv" )

# F) run a third simulation with dropping learners and s=0 (asocial learning)
sim3 <- sensitivity_NBDA_ind_error(x=together_apart_output, # with s=0 and dropping learners
                                   sightings=sightings,
                                   cutoff=cutoff, 
                                   association_index="SRI", 
                                   iterations=10000, 
                                   s=0, 
                                   num_ind_learn=20, 
                                   cores=2,
                                   keep_learners = FALSE,
                                   delta_AICc = 2)
save(sim3, file="sensitivity_NBDA_with_error_false pos_drop learners.RData") # save the sim3 object 
write.csv(sim3$raw,"sensitivity_NBDA_with_error_false pos_drop learners_RAW.csv" )
write.csv(sim3$summary,"sensitivity_NBDA_with_error_false pos_drop learners_SUMMARY.csv" )

# G) run a fourth simulation with keeping learners and s=0 (asocial learning)
sim4 <- sensitivity_NBDA_ind_error(x=together_apart_output, # with s=0 and keeping learners
                                   sightings=sightings,
                                   cutoff=cutoff, 
                                   association_index="SRI", 
                                   iterations=10000, 
                                   s=0, 
                                   num_ind_learn=20, 
                                   cores=2,
                                   keep_learners = TRUE,
                                   delta_AICc = 2)
save(sim4, file="sensitivity_NBDA_with_error_false pos_keep learners.RData") # save the sim4 object 
write.csv(sim4$raw,"sensitivity_NBDA_with_error_false pos_keep learners_RAW.csv" )
write.csv(sim4$summary,"sensitivity_NBDA_with_error_false pos_keep learners_SUMMARY.csv" )

#########################################################################################################


## plot results of sim1 

windowsFonts(A = windowsFont("Times New Roman")) 

par(mfrow=c(1,2))

par(mar=c(5.1,4,3,1)) # set figure margins
plot(sim1$summary[,"cutoff"], sim1$summary[,"perc_social_learning_supported"], # plot the proportion of social models supported against the cut-off points. 
     type="b",
     ylim=c(0,100),
     xlab="Cut-off points", 
     ylab="Percent",
     family="A",
     main="a) dropping learners",
     cex.main=1)

par(new=TRUE) # prevent R from opening a new figure

plot(sim1$summary[,"cutoff"], sim1$summary[,"within_95%_CI"], # plot estimates for s falling within 95% CI into the same plot
     type="b",
     xaxt="n", 
     yaxt="n",
     xlab="",
     ylab="",
     col="red",
     pch=2,
     ylim=c(0,100),
     family="A") # set limits of y-axis to match your minimum and maximum estimates

par(family="serif")
legend(x=1.5,y=20, pch=c(1,2), legend=c("Social models supported","s withing 95% C.I."), # add a legend
       lty=c(1,4), col=c("black","red"),
       bty="n", 
       cex=0.9)




#######################################################################################################
## plot results of sim2
par(mar=c(5.1,2,3,3)) # set figure margins
plot(sim2$summary[,"cutoff"], sim2$summary[,"perc_social_learning_supported"], # plot the proportion of social models supported against the cut-off points. 
     type="b",
     ylim=c(0,100),
     xlab="Cut-off points", 
     ylab="",
     main="b) retaining learners",
     cex.main=1)

par(new=TRUE) # prevent R from opening a new figure

plot(sim2$summary[,"cutoff"], sim2$summary[,"within_95%_CI"], # plot estimates of s falling within 95% CI into the same plot 
     type="b",
     xaxt="n", 
     yaxt="n",
     xlab="",
     ylab="",
     col="red",
     pch=2,
     ylim=c(0,100)) # set limits of y-axis to match your minimum and maximum estimates 


par(family="serif")
legend(x=1.5,y=20, pch=c(1,2), legend=c("Social models supported", "s within 95% C.I."), # add a legend
       lty=c(1,4), col=c("black","red"),
       bty="n", 
       cex=0.9)



########################################################################################################
## plot results of sim3
windowsFonts(A = windowsFont("Times New Roman")) 

par(mfrow=c(1,2))

par(mar=c(5.1,4,3,1)) # set figure margins
plot(sim3$summary[,"cutoff"], sim3$summary[,"perc_social_learning_supported"], # plot the proportion of social models supported against the cut-off points. 
     type="b",
     ylim=c(0,100),
     xlab="Cut-off points", 
     ylab="Percent",
     family="A",
     main="a) dropping learners",
     cex.main=1)

par(new=TRUE) # prevent R from opening a new figure

plot(sim3$summary[,"cutoff"], sim3$summary[,"within_95%_CI"], # plot estimates of s falling within 95% CI into the same plot 
     type="b",
     xaxt="n", 
     yaxt="n",
     xlab="",
     ylab="",
     col="red",
     pch=2,
     ylim=c(0,100),
     family="A") # set limits of y-axis to match your minimum and maximum estimates 

par(family="serif")
legend(x=1.5,y=50, pch=c(1,2), legend=c("Social models supported", "s within 95% C.I."), # add a legend
       lty=c(1,4), col=c("black","red"),
       bty="n", 
       cex=0.9)

#########################################################################################################
# plot results of sim4
par(mar=c(5.1,2,3,3)) # set figure margins
plot(sim4$summary[,"cutoff"], sim4$summary[,"perc_social_learning_supported"], # plot the proportion of social models supported against the cut-off points. 
     type="b",
     ylim=c(0,100),
     xlab="Cut-off points", 
     ylab="Percent",
     family="A",
     main="b) retaining learners",
     cex.main=1)

par(new=TRUE) # prevent R from opening a new figure

plot(sim4$summary[,"cutoff"], sim4$summary[,"within_95%_CI"], # plot estimates of s falling within 95% CI into the same plot 
     type="b",
     xaxt="n", 
     yaxt="n",
     xlab="",
     ylab="",
     col="red",
     pch=2,
     ylim=c(0,100),
     family="A") # set limits of y-axis to match your minimum and maximum estimates 

par(family="serif")
legend(x=1.5,y=50, pch=c(1,2), legend=c("Social models supported", "s within 95% C.I."), # add a legend
       lty=c(1,4), col=c("black","red"),
       bty="n", 
       cex=0.9)
########################################################################################################