#This software is in the public domain because it contains materials that originally came from the U.S. Geological Survey, an agency of the #United States Department of Interior. For more information, see the official USGS copyright policy

#Although this software program has been used by the U.S. Geological Survey (USGS), no warranty, expressed or implied, is made by the USGS #or the U.S. Government as to the accuracy and functioning of the program and related program material nor shall the fact of distribution #constitute any such warranty, and no responsibility is assumed by the USGS in connection therewith.

#This software is provided "AS IS."




###The Libraries
require(ggplot2)
require(colorspace)
require(gridExtra)
require(ggthemes)
require(RColorBrewer)
require(dplyr)
require(reshape)
require(tibble)
require(pls)
require(calibrate)
require(plsdepot)
require(MuMIn)
require(AICcmodavg)
require(corrplot)
require(sinkr)
require(tidyverse)
require(earlywarnings)
require(nlme)
require(car)
require(sem)
require(ape) 
require(caper) 
require(nlme) 
require(lavaan) 
require(piecewiseSEM)
require(corrgram)
require(relations)
require(ellipse)
require(QuantPsyc)
require(effects)
require(semPlot)



######The Data
Threats  <- read.csv("//Igsarfebfslacr3/Users/wthogmartin/My Documents/wet_migration/Monarch/Threats/FullThreatsData.csv")

plot(Threats$year, Threats$area_pva, xlab="Year", ylab="log(Population Size (ha))", cex=3, ylim=c(0, 2.8))



##data preparation, imputing missing values, example, not all imputations shown here
MBBRmatrix <- as.matrix(Threats[,64:67])
# dput(MBBRmatrix)
# dim(MBBRmatrix)

pal <- colorRampPalette(c("blue", "cyan", "yellow", "red"))
image(scale(MBBRmatrix), col=pal(100))
result_dineof <- dineof(scale(MBBRmatrix))
Xa <- result_dineof$Xa
image(Xa, col=pal(100))
MBBRmatrix[22,2]
sweep(result_dineof$Xa,MARGIN=2,attr(result_dineof$Xa, 'scaled:scale'),`*`) + sweep(result_dineof$Xa,MARGIN=2,attr(result_dineof$Xa, 'scaled:center'),"+")

image(MBBRmatrix, col=pal(100))
pred <- sweep(result_dineof$Xa,MARGIN=2,attr(result_dineof$Xa, 'scaled:scale'),`*`) + sweep(result_dineof$Xa,MARGIN=2,attr(result_dineof$Xa, 'scaled:center'),"+")
image(pred, col=pal(100))



###If interested in global climate processes
ENSO_DJF <- c(0.1, 0.9, -0.9, -0.5, 2.1, -1.4, -1.6, -0.7, -0.2, 0.9, 0.3, 0.7, -0.7, 0.7, -1.4, -0.7, 1.3, -1.3, -0.7, -0.4, -0.5, 0.6)  #The first in the series is Dec '93 and Jan-Feb '94
ENSO_NDJ <- c(0.1, 1.0, -0.9, -0.5, 2.3, -1.4, -1.6, -0.8, -0.3, 1.1, 0.3, 0.7, -0.7, 0.9, -1.3, -0.7, 1.3, -1.4, -0.9, -0.2, -0.3, 0.6)  #Data from http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml

ENSO <- as.data.frame(cbind(1993:2014, ENSO_DJF, ENSO_NDJ))
colnames(ENSO)[1] <- "year"

TolucaWeather <- merge(Threats[,68-79], ENSO, by="year", all.x=TRUE)
str(TolucaWeather)
MexW <- cor(TolucaWeather, use="pairwise.complete.obs")
corrplot(MexW, type="upper", tl.cex = 0.75, tl.col="black", tl.srt = 45)

ggplot(TolucaWeather,aes(x = year,y = OWmean_min_temp)) + xlim(1993,2014) + geom_point(colour="blue") + geom_path(colour="blue") + ylab("Mean minimum temperature (C)")


Threats <- merge(Threats,TolucaWeather,by="year")
str(Threats)
Threats <- subset(Threats, select=-c(Month,Day))


########### Start from here given the data associated with the paper
Threats <- FullThreats  <- read.csv("//Igsarfebfslacr3/Users/wthogmartin/My Documents/wet_migration/Monarch/Threats/FullThreatsData.csv")

plot(Threats$year, Threats$area_pva, xlab="Year", ylab="log(Population Size (ha))", cex=3, ylim=c(0, 2.8))




Threatscorr <- cor(Threats)  #with diff
#corrplot(Threatscorr, method="ellipse")
corrplot(Threatscorr[c(2,64:81),c(2,64:81)], method="ellipse")
corrplot(Threatscorr[c(2,3:15),c(2,3:15)], method="ellipse")
corrplot(Threatscorr[c(2,16:25),c(2,16:25)], method="ellipse")
corrplot(Threatscorr[c(2,26:41),c(2,26:41)], method="ellipse")
corrplot(Threatscorr[c(2,42:52),c(2,42:52)], method="ellipse")
corrplot(Threatscorr[c(2,53:63),c(2,53:63)], method="ellipse")



#############Partial Least Squares, Population Size
head(Threats)
str(Threats)
ThreatsOrdered2 = Threats[ ,c(3,5:79, 2)]
str(ThreatsOrdered2)
#########

pls10 = plsreg1(ThreatsOrdered2[, 1:76], ThreatsOrdered2[, 77, drop = FALSE], crosval = TRUE)


pls10$R2
plot(pls10, ylab="Component 2", xlab="Component 1", main="Trend")
sum(pls10$R2)

pls10$x.scores     

pls10$reg.coefs

pls10$Q2

pls10$x.scores          # X-scores (T-components)
pls10$x.loads           # X-loadings
pls10$y.scores          # Y-scores (U-components)
pls10$y.loads           # Y-loadings
pls10$cor.xyt           #  score correlations
pls10$raw.wgs           # raw weights
pls10$mod.wgs           # modified weights
pls10$std.coefs         # standard coefficients
pls10$reg.coefs         # regular coefficients
sum(pls10$R2)                #  R-squared
pls10$R2
pls10$R2[1]/pls10$R2[2]
pls10$R2Xy              #  explained variance of X-y by T
pls10$y.pred            #  y-predicted
pls10$resid             #  residuals
pls10$T2                #  T2 hotelling
pls10$Q2                #  Q2 cross validation; PRESS predicted residual sum of squares
pls10$Q2[1,3]+pls10$Q2[2,3]


pls10$x.loads[
  order( pls10$x.loads[,1], pls10$x.loads[,2] ),
  ]

pls10$x.loads[
  order( pls10$x.loads[,2], pls10$x.loads[,1] ),
  ]

plot(Threats$year, Threats$area_pva, xlab="Year", ylab="log(Population Size (ha))", cex=3, ylim=c(0, 2.8))
points(Threats$year, pls10$y.pred, cex=2, pch=19) #pls10$y.pred calculated below


# Prepare file for export and minor manipulation in Excel
# y2yloads10 <- data.frame(pls10$x.loads)
# y2yloads10 <- rownames_to_column(y2yloads10, "VALUE")
# colnames(y2yloads10)[1] <- "Variable"
# y2yloadsmelt10 <- melt(y2yloads10)
# colnames(y2yloadsmelt10)[2] <- "Component"
# write.table(y2yloadsmelt10, file="//.../y2yloadsmelt10.csv")

#Import component loadings file
component_loadings  <- read.csv("//Igsarfebfslacr3/Users/wthogmartin/My Documents/wet_migration/Monarch/Threats/component_loadings.csv")   ####Added, in Excel, groups; winnowed to set of variables with >|0.15| in p1 or p2

component_loadings$Component <- as.character(component_loadings$Component)
component_loadings$Component <- replace(component_loadings$Component, component_loadings$Component=="p1", "Component 1")
component_loadings$Component <- replace(component_loadings$Component, component_loadings$Component=="p2", "Component 2")


color_palette <- colorRampPalette(c("#3794bf", "#FFFFFF", "#df8640"))

ggplot(component_loadings, aes(y=reorder(Variable, ID), x=Loading, fill=Component)) + 
  geom_vline(xintercept=0, linetype=2, size=2, color="red") + 
  geom_point(aes(colour = Group, size=abs(Loading))) +
  theme_light() + 
  scale_colour_brewer(type = "qual", palette = 3, direction = -1) + 
  #aes(y=reorder(Variable, abs(value))
  theme(axis.text.x = element_text(angle=90, size = 14, vjust=0.5),
        legend.background = element_rect(fill = "white"),
        axis.title.y=element_text(margin=margin(0,20,0,0), size = 20), 
        axis.title.x=element_text(margin=margin(20,0,0,0), size = 20),
        legend.text=element_text(size=15),
        strip.text.x = element_text(size=16),
        legend.key = element_blank(),
        plot.title = element_text(hjust = 0, vjust=2.12, size=20) ) +
  facet_grid(.~Component) +
  labs(y="Variable", x="Loading", size="Magnitude", colour="Type") +  
  guides(fill=FALSE, colour = guide_legend(override.aes = list(size=5))) +
  ggtitle("Annual Population Size (ha)")



####circle of correlations
explvar(pls10)
plot(pls10, plottype = "scores", comps = 1:3, cex=0.5)





###Select subset with loadings |>0.15|
plot(pls10$x.loads[,1], cex=2)
text(pls10$x.loads[,1], rownames(pls10$x.loads), cex=0.6, pos=4, col="red")
abline(h = c(-0.15, 0.15))

plot(pls10$x.loads[,2], cex=2)
text(pls10$x.loads[,2], rownames(pls10$x.loads), cex=0.6, pos=4, col="blue")
abline(h = c(-0.15, 0.15))


#######Relative Importance
# require(AICcmodavg)
# require(MuMIn)

cor(Threats$glycum_N, Threats$glycum_S)
cor(Threats$totalneon_N, Threats$totalneon_S)
cor(Threats$TminJu_NC, Threats$TminJl_NC)
cor(Threats$T70p4sum_NE, Threats$Tempp4avg_NE)
cor(Threats$T70p4sum_NC, Threats$T70p4sum_NE)
cor(Threats$Tempp4avg_NE, Threats$Tempp4avg_NC)
cor(Threats$Ramirez.cumul, Threats$Vidal.cumul)

BestSubset <- as.data.frame(cbind(Threats$Ramirez.cumul,
                  Threats$glycum_N,
                  Threats$totalneon_N,
                  Threats$DC_S,
                  Threats$apva_1yr,
#                  Threats$Tempp4avg_NE,  #Basically the same variable as T70p4sum_NE
                  Threats$GDD_NC,
                  Threats$T70p1sum_NC,
                  Threats$T70p4sum_NE,
                  Threats$TminAug_NC,
                  Threats$MeanJ_NC,
                  Threats$TminJu_NC))
BestSubsetcorr <- cor(BestSubset)  #with diff
corrplot(BestSubsetcorr, method="ellipse")

BestSubsetscale <- data.frame(lapply(BestSubset, function(x) scale(x, center = TRUE, scale = TRUE)))
colnames(BestSubsetscale) <- c("Ramirez.cumul" ,
                                 "glycum_N" ,
                                 "totalneon_N" ,
                                 "DC_S" ,
                                 "apva_1yr" ,
                                 "GDD_NC" ,
                                 "T70p1sum_NC" ,
                                 "T70p4sum_NE" ,
                                 "TminAug_NC" ,
                                 "MeanJ_NC" ,
                                 "TminJu_NC")
BestSubsetscale["year"] <- c(1993:2014)
str(BestSubsetscale)

BestSubsetscale["area_pva"] <- Threats$area_pva

GlobMod.FullNorth <- gls(area_pva ~ 
                         Ramirez.cumul +
                         glycum_N +
                         totalneon_N +
                         DC_S +
                         apva_1yr +
                         GDD_NC +
                         T70p1sum_NC +
                         T70p4sum_NE +
                         TminAug_NC +
                         MeanJ_NC +
                         TminJu_NC, na.action = na.fail, 
                         correlation=corAR1(form= ~year), data=BestSubsetscale)
acf(GlobMod.FullNorth$res) 
summary(GlobMod.FullNorth)

GlobMod.FullNorthLM <- lm(area_pva ~ 
                            Ramirez.cumul +
                            glycum_N +
                            totalneon_N +
                            DC_S +
                            apva_1yr +
                            GDD_NC +
                            T70p1sum_NC +
                            T70p4sum_NE +
                            TminAug_NC +
                            MeanJ_NC +
                            TminJu_NC, na.action = na.fail, 
                          data=BestSubsetscale)
acf(GlobMod.FullNorthLM$res) 

autoregressivetest <- gls(area_pva ~ TminAug_NC , na.action = na.fail, 
                         correlation=corAR1(form= ~year), data=Threats)



ModSetRes.FullNorthLM <- dredge(GlobMod.FullNorthLM, m.lim=c(0,3), extra = "R^2", rank = "BIC")
AvgModSetRes.FullNorthLM <- model.avg(ModSetRes.FullNorthLM)


AvgModSetRes.FullNorthLM$coefficients

varlistNorth <- c(  
  "Ramirez.cumul" ,
    "glycum_N" ,
    "totalneon_N" ,
    "DC_S" ,
    "apva_1yr" ,
    "GDD_NC" ,
    "T70p1sum_NC" ,
    "T70p4sum_NE" ,
    "TminAug_NC" ,
    "MeanJ_NC" ,
    "TminJu_NC")
sort(varlistNorth)

options(scipen=999)

wt.Full_NorthLM <- summary(AvgModSetRes.FullNorthLM)$importance[c("Ramirez.cumul" ,
                                                                     "glycum_N" ,
                                                                     "totalneon_N" ,
                                                                     "DC_S" ,
                                                                     "apva_1yr" ,
                                                                     "GDD_NC" ,
                                                                     "T70p1sum_NC" ,
                                                                     "T70p4sum_NE" ,
                                                                     "TminAug_NC" ,
                                                                     "MeanJ_NC" ,
                                                                     "TminJu_NC")]	




subset(ModSetRes.FullNorthLM, delta <2)
subset(ModSetRes.FullNorthLM, 1/8 < weight/max(ModSetRes.FullNorthLM$weight))
subset(ModSetRes.FullNorthLM, cumsum(ModSetRes.FullNorthLM$weight) <= .95) 
ModSetRes.FullNorthLM[1:3]



pred.parms <- get.models(ModSetRes.FullNorthLM, subset= delta < 2) 
# predict values using each model, here were just using the 
# the example dataset, you could use a new dataset 
model.preds = sapply(pred.parms, predict, newdata = Threats)

ModSetRes.FullNorthLM[1]
bestmodel <- lm(area_pva ~ glycum_N + T70p1sum_NC + 
                  TminAug_NC, data = Threats, na.action = na.fail)
summary(bestmodel)
confint(bestmodel)

outlierTest(bestmodel) # Bonferonni p-value for most extreme obs
qqPlot(bestmodel, main="QQ Plot") #qq plot for studentized resid 
leveragePlots(bestmodel) 



#############Confounding
Confounding <- lm(area_pva ~ 
                            Ramirez.cumul +
                            glycum_N +
                            totalneon_N +
                            DC_S +
                            apva_1yr +
                            T70p1sum_NC +
                            TminAug_NC, na.action = na.fail, 
                          data=BestSubsetscale)

Confounding.dredge <- dredge(Confounding, m.lim=c(0,3), extra = "R^2", rank = "BIC")
Confounding.average <- model.avg(Confounding.dredge)

Confound.overwinter <- lm(area_pva ~ 
                    Ramirez.cumul +
                    glycum_N +
                    T70p1sum_NC +
                    TminAug_NC, na.action = na.fail, 
                  data=Threats)
Confound.neon <- lm(area_pva ~ 
                    glycum_N +
                    totalneon_N +
                    T70p1sum_NC +
                    TminAug_NC, na.action = na.fail, 
                  data=Threats)
Confound.dicamba <- lm(area_pva ~ 
                    glycum_N +
                    DC_S +
                    T70p1sum_NC +
                    TminAug_NC, na.action = na.fail, 
                  data=Threats)
Confound.density <- lm(area_pva ~ 
                    glycum_N +
                    apva_1yr +
                    T70p1sum_NC +
                    TminAug_NC, na.action = na.fail, 
                  data=Threats)


Confound.MT70P1 <- lm(area_pva ~ 
                         glycum_N +

                         TminAug_NC, na.action = na.fail, 
                       data=Threats)
Confound.MTminAug <- lm(area_pva ~ 
                         glycum_N +
                         T70p1sum_NC, na.action = na.fail, 
                       data=Threats)




con.overwinter <- lm(area_pva ~ 
                            Ramirez.cumul +
                            T70p1sum_NC +
                            TminAug_NC, na.action = na.fail, 
                          data=Threats)
con.neon <- lm(area_pva ~ 
                      totalneon_N +
                      T70p1sum_NC +
                      TminAug_NC, na.action = na.fail, 
                    data=Threats)
con.dicamba <- lm(area_pva ~ 
                         DC_S +
                         T70p1sum_NC +
                         TminAug_NC, na.action = na.fail, 
                       data=Threats)
con.density <- lm(area_pva ~ 
                         apva_1yr +
                         T70p1sum_NC +
                         TminAug_NC, na.action = na.fail, 
                       data=Threats)


summary(bestmodel)
summary(Confound.overwinter)
summary(Confound.neon)
summary(Confound.dicamba)
summary(Confound.density)

beta <- as.data.frame(c(bestmodel$coefficients[2], 
          Confound.overwinter$coefficients[3], 
          Confound.neon$coefficients[2], 
          Confound.dicamba$coefficients[2], 
          Confound.density$coefficients[2]))
rownames(beta) <- c("Best Model","w/ Overwinter\nForest Loss","w/ Neonicotinoids","w/ Dicamba", "w/ Previous Year's\n  Abundance")
colnames(beta) <- "slope"

ggplot(mapping=aes(x="", y=beta$slope, label=rownames(beta))) + theme_bw() + 
  geom_point(size=3)+geom_text(hjust=0, vjust=1.1) +
  xlab("Model") + 
  ylab("Slope") + theme(panel.border = element_blank(),aspect.ratio=6/3)
 
(bestmodel$coefficients[2]-Confound.overwinter$coefficients[3])/bestmodel$coefficients[2]
(bestmodel$coefficients[2]-Confound.neon$coefficients[2])/bestmodel$coefficients[2]
(bestmodel$coefficients[2]-Confound.dicamba$coefficients[2])/bestmodel$coefficients[2]
(bestmodel$coefficients[2]-Confound.density$coefficients[2])/bestmodel$coefficients[2]

(con.overwinter$coefficients[2]-Confound.overwinter$coefficients[2])/con.overwinter$coefficients[2]
(con.neon$coefficients[2]-Confound.neon$coefficients[3])/con.neon$coefficients[2]
(con.dicamba$coefficients[2]-Confound.dicamba$coefficients[3])/con.dicamba$coefficients[2]
(con.density$coefficients[2]-Confound.density$coefficients[3])/con.density$coefficients[2]

con.year <- lm(area_pva ~ year + T70p1sum_NC + 
                  TminAug_NC, data = Threats, na.action = na.fail)
summary(con.year)

plot(resid(con.year)~Threats$glycum_N)
abline(lm(resid(con.year)~Threats$glycum_N), col="red") # regression line (y~x) 
lines(lowess(resid(con.year),Threats$glycum_N), col="blue") # lowess line (x,y)


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

# Influential Observations
# Cook's D plot
# identify D values > 4/(n-k-1) 
cutoff <- 4/((nrow(Threats)-length(bestmodel$coefficients)-2)) 
plot(bestmodel, which=4, cook.levels=cutoff)
# Influence Plot 
#influencePlot(bestmodel,	id.method="identify", main="Influence Plot", sub="Circle size is proportial to Cook's Distance" )

bestpred.parms <- get.models(ModSetRes.FullNorthLM, subset= delta < 2) 

bestmodel.preds = sapply(bestpred.parms, predict, newdata = BestSubsetscale)

####Figure 4
plot(0,0,xlim = c(1992,2015),ylim = c(0,2.7), xlab="Year", ylab="log(Population Size (ha))")
points(Threats$year, bestmodel.preds, cex=2, pch=21, col="red", bg="red") #Reduced Model 1
# points(Threats$year, model.preds[,2], cex=2, pch=24, col="green", bg="green", type="o") #Reduced Model 2
# points(Threats$year, model.preds[,3], cex=2, pch=25, col="tan", bg="tan", type="o")  #Reduced Model 3
points(Threats$year, pls10$y.pred, cex=2, pch=19)  #PLSR predicted
points(Threats$year, Threats$area_pva, cex=3, type="o") #Observed
points(2005, 2.45, cex=3, type="o")
points(2005, 2.3, cex=2, pch=19)
points(2005, 2.15, cex=2, pch=21, col="red", bg="red")
text(2007.12, 2.45, "Observed")
text(2010.7, 2.3, "Partial Least Squares Prediction")
text(2010.28, 2.15, "Best Subset Model Prediction")


cor(Threats$area_pva,bestmodel.preds)
cor(Threats$area_pva,pls10$y.pred)


Threats$area_pva - bestmodel.preds
sum(Threats$area_pva - bestmodel.preds)
max(abs(Threats$area_pva - bestmodel.preds))

Threats$area_pva - pls10$y.pred
sum(Threats$area_pva - pls10$y.pred)
max(abs(Threats$area_pva - pls10$y.pred))




obspred <- cbind(Threats$area_pva, Threats$year, bestmodel.preds, pls10$y.pred)
colnames(obspred) <- c("Observed", "Year", "Reduced 1", "Full 1")
cormatrix1 <- cor(obspred)


corrgram(obspred)
corrplot(cormatrix1, method="ellipse")


plotcorr(cormatrix1,  type = "lower")

cor.mtest <- function(mat, ...) {
  mat <- as.matrix(mat)
  n <- ncol(mat)
  p.mat<- matrix(NA, n, n)
  diag(p.mat) <- 0
  for (i in 1:(n - 1)) {
    for (j in (i + 1):n) {
      tmp <- cor.test(mat[, i], mat[, j], ...)
      p.mat[i, j] <- p.mat[j, i] <- tmp$p.value
    }
  }
  colnames(p.mat) <- rownames(p.mat) <- colnames(mat)
  p.mat
}
# matrix of the p-value of the correlation
p.mat <- cor.mtest(obspred)
head(p.mat[, 1:4])



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


coef_lmbeta <- lm.beta(bestmodel)
BIC(bestmodel)

gls1 <- gls(area_pva ~ glycum_N + T70p1sum_NC + 
              TminAug_NC, 
            correlation=corAR1(form= ~year), data=Threats)

summary(gls1)
intervals(gls1)
BIC(gls1)
acf(bestmodel$res)




efflm1 <- allEffects(bestmodel)
efflm1GLY <- effect("glycum_N", bestmodel)
efflm1T70 <- effect("T70p1sum_NC", bestmodel)
efflm1Tmin <- effect("TminAug_NC", bestmodel)


plot(allEffects(bestmodel),"glycum_N",
     xlab="Cumulative Total Glyphosate Application (kg)", 
     ylab="log(Winter Population Size (ha))",
     main="", rug=FALSE)
plot(allEffects(bestmodel),"T70p1sum_NC",
     xlab="Total Number of Days in Period 1 >21"~degree~C, 
     ylab="log(Winter Population Size (ha))",
     main="", rug=FALSE)
plot(allEffects(bestmodel),"TminAug_NC",
     xlab="August Minimum Temperature"~degree~C, 
     ylab="log(Winter Population Size (ha))",
     main="", rug=FALSE)





fancy_scientific <- function(l) {
  # turn in to character string in scientific notation
  l <- format(l, scientific = TRUE)
  # quote the part before the exponent to keep all the digits
  l <- gsub("^(.*)e", "'\\1'e", l)
  # turn the 'e+' into plotmath format
  l <- gsub("e", "%*%10^", l)
  # return this as an expression
  parse(text=l)
}


## Change effect object to dataframe
efflm1GLY_df <- data.frame(efflm1GLY)
efflm1T70_df <- data.frame(efflm1T70)
efflm1Tmin_df <- data.frame(efflm1Tmin)



## Figure 3

ggplot(data=efflm1GLY_df) + geom_line(aes(glycum_N,fit), size=2, color="black") + theme_bw() +
  xlab("Cumulative Total Glyphosate Application (kg)") + 
  ylab("log(Winter Population Size (ha))") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom") + 
  geom_ribbon(aes(ymin=lower, ymax=upper, x=glycum_N), alpha=0.3) + 
  scale_x_continuous(labels=fancy_scientific)

ggplot(data=efflm1T70_df) + geom_line(aes(T70p1sum_NC,fit), size=2, color="black") + theme_bw()+
  xlab(expression("Total Number of Days in Period 1 >21"*~degree*"C")) + 
  ylab("log(Winter Population Size (ha))") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom") +
  geom_ribbon(aes(ymin=lower, ymax=upper, x=T70p1sum_NC), alpha=0.3) + 
  scale_x_continuous(labels=fancy_scientific)


ggplot(efflm1Tmin_df) + geom_line(aes(TminAug_NC,fit), size=2, color="black") + theme_bw()+
  xlab(expression("August Minimum Temperature"*~degree*"C")) + 
  ylab("log(Winter Population Size (ha))") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom") +
  geom_ribbon(aes(ymin=lower, ymax=upper, x=TminAug_NC), alpha=0.3)  
  

####With lines instead of ribbons
ggplot(data=efflm1GLY_df) + geom_line(aes(glycum_N,fit), size=2, color="black") + theme_bw() +
  xlab("Cumulative Total Glyphosate Application (kg)") + 
  ylab("log(Winter Population Size (ha))") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom") + 
  geom_line(aes(glycum_N,lower),linetype=1.5, size=2, color="dark gray") + 
  geom_line(aes(glycum_N,upper),linetype=1.5, size=2, color="dark gray") +
  scale_x_continuous(labels=fancy_scientific)

ggplot(data=efflm1T70_df) + geom_line(aes(T70p1sum_NC,fit), size=2, color="black") + theme_bw()+
  xlab(expression("Total Number of Days in Period 1 >21"*~degree*"C")) + 
  ylab("log(Winter Population Size (ha))") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom") +
  geom_line(aes(T70p1sum_NC,lower),linetype=1.5, size=2, color="dark gray") + 
  geom_line(aes(T70p1sum_NC,upper),linetype=1.5, size=2, color="dark gray") +
  scale_x_continuous(labels=fancy_scientific)


ggplot(efflm1Tmin_df) + geom_line(aes(TminAug_NC,fit), size=2, color="black") + theme_bw()+
  xlab(expression("August Minimum Temperature"*~degree*"C")) + 
  ylab("log(Winter Population Size (ha))") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom") +
  geom_line(aes(TminAug_NC,lower),linetype=1.5, size=2, color="dark gray") + 
  geom_line(aes(TminAug_NC,upper),linetype=1.5, size=2, color="dark gray") 
  





require(scales)
require(grid)
require(gridExtra)
require(gtable)
require(cowplot)

Figa <- ggplot(efflm1GLY_df)+geom_line(aes(glycum_N,fit),size=2, color="black") + theme_bw()+
  xlab("Cumulative Total\nGlyphosate Application (kg)") + 
  ylab("Winter Population Size (ha)") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom",plot.margin=unit(c(1,1,2,0),"cm")) +
  geom_line(aes(glycum_N,lower),linetype=1.5, size=2, color="dark gray") + 
  geom_line(aes(glycum_N,upper),linetype=1.5, size=2, color="dark gray") +
  scale_x_continuous(labels=fancy_scientific)


Figb <- ggplot(efflm1T70_df)+geom_line(aes(T70p1sum_NC,fit),size=2, color="black") + theme_bw()+
  xlab(expression("Total Number of Days\nin Period 1 >21"*~degree*"C")) + 
  ylab("") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom",plot.margin=unit(c(1,1,2,0),"cm"),axis.title.x = element_text(vjust=-0.5)) +
  geom_line(aes(T70p1sum_NC,lower),linetype=1.5, size=2, color="dark gray") + 
  geom_line(aes(T70p1sum_NC,upper),linetype=1.5, size=2, color="dark gray") +
  scale_x_continuous(labels=fancy_scientific)


Figc <- ggplot(efflm1Tmin_df)+geom_line(aes(TminAug_NC,fit),size=2, color="black") + theme_bw()+
  xlab(expression("August Minimum\nTemperature"*~degree*"C")) + 
  ylab("") + 
  coord_cartesian(ylim=c(0, 2.5)) +
  theme(legend.position="bottom",plot.margin=unit(c(1,1,2,0),"cm")) +
  geom_line(aes(TminAug_NC,lower),linetype=1.5, size=2, color="dark gray") + 
  geom_line(aes(TminAug_NC,upper),linetype=1.5, size=2, color="dark gray") 



Figag <- ggplotGrob(Figa)
Figbg <- ggplotGrob(Figb)
Figcg <- ggplotGrob(Figc)


title.groba <- textGrob(
  label = "a",
  x = unit(0, "lines"), 
  y = unit(0, "lines"),
  hjust = -6, vjust = 0,
  gp = gpar(fontsize = 16))

p1a <- arrangeGrob(Figag, top = title.groba)
grid.arrange(p1a)

title.grobb <- textGrob(
  label = "b",
  x = unit(0, "lines"), 
  y = unit(0, "lines"),
  hjust = -6, vjust = 0,
  gp = gpar(fontsize = 16))

p1b <- arrangeGrob(Figbg, top = title.grobb)
grid.arrange(p1b)

title.grobc <- textGrob(
  label = "c",
  x = unit(0, "lines"), 
  y = unit(0, "lines"),
  hjust = -6, vjust = 0,
  gp = gpar(fontsize = 16))

p1c <- arrangeGrob(Figcg, top = title.grobc)
grid.arrange(p1c)

#plot_grid(Figa, Figb, Figc, align = "v", ncol = 3, rel_heights = c(1/3, 1/3, 1/3))
grid.arrange(p1a, p1b, p1c, ncol=3)




###############Replace glyphosate with milkweed availability
cor(x=Threats$glycum_N,y=Threats$milkweed.resource)
cor(x=Threats$glycum_N,y=Threats$log.milkweed.resource)
cor(x=Threats$glysum_N,y=Threats$milkweed.resource)
cor(x=Threats$glysum_N,y=Threats$log.milkweed.resource)
cor(x=Threats$year,y=Threats$milkweed.resource)
cor(x=Threats$year,y=Threats$log.milkweed.resource)
cor(x=Threats$glysum_N,y=Threats$area_pva)
cor(x=Threats$glysum_N,y=Threatsw$apva_1yr)
cor(y=Threats$area_pva,x=Threats$milkweed.resource)
cor(y=Threats$area_pva,x=Threats$log.milkweed.resource)



par(mfrow=c(2,2))
plot(log(Threats$glycum_N), log(Threats$milkweed.resource), ylab="log(Milkweed Resource)", xlab="log(Cumulative Glyphosate)", cex=3, cex.axis=1.5, cex.lab=1.5)

plot(Threats$year, log(Threats$milkweed.resource), xlab="Year", ylab="log(Milkweed Resource)", cex=3, cex.axis=1.5, cex.lab=1.5)

plot(x=log(Threats$milkweed.resource), y=Threats$area_pva, xlab="log(Milkweed Resource)", ylab="log(Overwinter Area (ha))", cex=3, cex.axis=1.5, cex.lab=1.5, ylim=c(0,3))
text(x=log(Threats$milkweed.resource), y=Threats$area_pva, labels=Threats$year, cex= 1, pos=3)

plot(Threats$year, log(Threats$glycum_N), xlab="Year", ylab="log(Cumulative Glyphosate)", cex=3, cex.axis=1.5, cex.lab=1.5, ylim=c(min(log(Threats$glycum_N)),max(log(Threats$glycum_N))))

par(mfrow=c(1,1))


#Structural Equation Model

semdata <- as.data.frame(cbind(Threats$area_pva,Threats$log.milkweed.resource,Threats$glycum_N))
colnames(semdata) <- c("area_pva", "milkweed.resource", "glycum_N")

mod.sem2 = 'area_pva ~ milkweed.resource
milkweed.resource ~ glycum_N'

#options(scipen=999)
options(scipen=0)

model.sem2 = sem(mod.sem2, semdata, estimator = "MLM")
parTable(model.sem2)
summary(model.sem2, standardize=TRUE)
summary(model.sem2,  fit.measures = TRUE)
parameterEstimates(model.sem2)
coef(model.sem2)
fitted(model.sem2)
summary(model.sem2, standardize=FALSE)
inspect(model.sem2, "rsquare")

semPaths(model.sem2, intercepts = FALSE)

mod.sem3 = 'area_pva ~ milkweed.resource + glycum_N
milkweed.resource ~ glycum_N'

model.sem3 = sem(mod.sem3, semdata, estimator = "MLM")
parTable(model.sem3)
summary(model.sem3, standardize=TRUE)
#summary(model.sem2, standardize=FALSE)
sem.model.fits(model.sem3)
inspect(model.sem3, "rsquare")

mod.sem4 = 'area_pva ~ milkweed.resource
milkweed.resource ~ glycum_N + glycum_N^2'

model.sem4 = sem(mod.sem4, semdata, estimator = "MLM")
parTable(model.sem4)
summary(model.sem4, standardize=TRUE)
summary(model.sem4,  fit.measures = TRUE)
parameterEstimates(model.sem4)
coef(model.sem4)
fitted(model.sem4)
summary(model.sem4, standardize=FALSE)
inspect(model.sem4, "rsquare")

semdata.l <- as.data.frame(cbind(Threats$area_pva,Threats$log.milkweed.resource,log(Threats$glycum_N)))
colnames(semdata.l) <- c("area_pva", "milkweed.resource", "glycum_N")


model.sem5 = sem(mod.sem2, semdata.l, estimator = "MLM")
parTable(model.sem5)
summary(model.sem5, standardize=TRUE)
summary(model.sem5,  fit.measures = TRUE)
parameterEstimates(model.sem5)
coef(model.sem5)
fitted(model.sem5)
summary(model.sem5, standardize=FALSE)
inspect(model.sem5, "rsquare")




mod.cfa2 <- cfa(mod.sem2, data=semdata)
parTable(mod.cfa2)
summary(mod.cfa2)


corsem <- cor(semdata)
plotcorr(corsem)

summary(lm(semdata$milkweed.resource~semdata$glycum_N))


semdata2 <- as.data.frame(cbind(Threats$area_pva,Threats$log.milkweed.resource,Threats$glycum_N, Threats$Ramirez.cumul))
colnames(semdata2) <- c("area_pva", "milkweed.resource", "glycum_N", "Ramirez.cumul")

mod2.sem = 'area_pva ~ milkweed.resource
milkweed.resource ~ glycum_N
area_pva ~ Ramirez.cumul'

model2.sem = sem(mod2.sem, semdata2, estimator = "MLM")
parTable(model2.sem)
summary(model2.sem, standardize=TRUE)
summary(model2.sem,  fit.measures = TRUE)
parameterEstimates(model2.sem)
coef(model2.sem)
fitted(model2.sem)
summary(model2.sem, standardize=FALSE)
inspect(model2.sem, "rsquare")

semPaths(model2.sem, intercepts = FALSE)