summary(auc_lmer)

summary(disease_lmer)

summary(psurv_lnorm)

cld(glht(auc_lmer, linfct=mcp(Strain="Tukey")))

#Compact letters for lmerTest objects are a little tricky. This solution comes from the rcompanion. Mangiafico, S.S. 2015. An R Companion for the Handbook of Biological Statistics, version 1.3.0. 
### Extract lsmeans table



cldList(comparison = Comparison,
        p.value    = p.adjust(lmerlsm$'p-value',
                       method =  "bonferroni")   ,
        threshold = 0.05)

cld(glht(srv_coxph,linfct=mcp(Strain="Tukey")))

cld(glht(psurv_lnorm,linfct=mcp(Strain="Tukey")))

cld(glht(cme,linfct=mcp(Strain="Tukey")))

sessionInfo()

####This script is used to generate  figures for the Manuscript####
#It requires that everything from the rmd file is available through the environment,
#in practical terms, that the markdown script was run through the R console.

#### Fig1 ####
##generate a visual display of 
#a) raw data.
#b) AUDPC 
#c) Linear fit and 
#d) survival curve

library("grid")

Fig1a <- ggplot(data=di_long) +
                  geom_jitter(aes(x=DPI, y=DI, color=Strain, shape=Batch)) +
                  #geom_segment(aes(x=0, xend=max(DPI)+0.5,y=cutoff, yend=cutoff), linetype="dotted") +
                  labs(x = "Days post infection", y = "Disease Index", title="a) Scatterplot of all disease indices") +
                  coord_cartesian(xlim=c(2,11))+
  theme_bw()
                  
Fig1b <-   ggplot(data=filter(di_summary, Batch=="B-I")) +
                  aes(x=DPI,y=mean, color=Strain,fill=Strain) +
                  geom_area(alpha=0.10, position="identity") +
                  labs(x = "Days post infection", y = "Avg. Disease Index", title="b) Area under the disease progression curve (Batch B-I)")+
  coord_cartesian(xlim=c(2,11))+
  theme_bw() #+ theme(legend.position="none")


Fig1c <- 
  ggplot(data=filter(di_long), aes(x=DPI,y=DI)) +
  geom_smooth(aes(y=lmer.pred,color=Strain),method="lm", se= F) +
  coord_cartesian(ylim=c(-0.5,4.5),xlim=c(2,11)) +
  #theme(legend.position="none") +
  labs(x = "Days post infection", y = "Disease Index", 
       title="c) Estimated disease development")+ #Labels
  theme_bw() 

Fig1d <-  ggplot(data=filter(surv_DI_fit.df),aes(x=time, y=surv, color=Strain)) + 
  geom_step(aes(linetype=Batch),alpha=1) + geom_line(data=sreg_curves,aes(y=lnorm.sreg ,color=Strain), size=1.1, alpha=0.6) +
  labs(x="Days post infection", y="Estimated survival", title="d) Kaplan-Meier survival estimates and fits to lognormal distribution") +
  coord_cartesian(xlim=c(2,11)) +
  theme_bw()

dev.off()
pushViewport(viewport(layout= grid.layout(nrow=2, ncol=2)))
print(Fig1a, vp = viewport(layout.pos.row = 1, layout.pos.col = 1))
print(Fig1b, vp = viewport(layout.pos.row = 1, layout.pos.col = 2))
print(Fig1c, vp = viewport(layout.pos.row = 2, layout.pos.col = 1))
print(Fig1d, vp = viewport(layout.pos.row = 2, layout.pos.col = 2))


###Figure 1 done####

####clear device
dev.off()

#### Fig 2 #####
##Figure on AUDPC
###For the purpose of visualization calculate summaries across all replicates
di_summary.nobatch <- di_long %>% group_by(Plant, Strain, DPI) %>% summarise(mean(DI),sd(DI),sd(DI)/sqrt(length(DI)))
colnames(di_summary.nobatch) <- c("Plant", "Strain", "DPI", "mean", "sd", "se")
####Fig2a: Plot the actual areas per strain
Fig2a <- ggplot(di_summary.nobatch) +
  aes(x=DPI,y=mean,color=Strain) +
  geom_area(aes(fill=Strain), position="identity",alpha=0.15) +
  geom_errorbar(aes(x=DPI, ymax=mean+se,ymin=mean-se))+
  facet_wrap(~Strain) +
  labs(x = "Days post infection", y = "Avg. Disease Index") +
  theme(legend.position="none") +
  ggtitle("a) Disease progression curves per strain")+
  theme_bw()

###Some extra scripting to make the mean and CI plot.
auc_CI <- as.data.frame(tidy(confint(auc_lmer)))
auc_CI <- auc_CI[3:9,] ##Drop sig01, sigma
auc_CI$Strain <- levels(auc_df$Strain)
###Mean relative to Strain1 (except strain1 that one is absolute)

for (i in 1:nrow(auc_CI)){
  if (i==1){
    auc_CI$mean[i] <- mean(c(auc_CI$X2.5..[i],auc_CI$X97.5..[i]))
    auc_CI$upr[i] <- auc_CI$X97.5..[i]
    auc_CI$lwr[i] <- auc_CI$X2.5..[i]
  } else {
    auc_CI$mean[i] <- c(auc_CI$mean[1]+mean(c(auc_CI$X2.5..[i],auc_CI$X97.5..[i])))
    auc_CI$upr[i] <- c(auc_CI$mean[1]+auc_CI$X97.5..[i])
    auc_CI$lwr[i] <- c(auc_CI$mean[1]+auc_CI$X2.5..[i])
  }
}

#####Fig2b meanCI the areas, and add each individual area as jitter; symbols in jitter by batch
Fig2b <- ####Generate meanCI of AUDCP with significance letters
  ggplot(aes(x=Strain, y=AUC, color=Strain),data=auc_df) + ###Plot the auc_df
  geom_crossbar(data = auc_CI, aes(x = Strain, y = mean, ymin = lwr, ymax = upr,fill=Strain), alpha=0.3) +
  geom_jitter(aes(shape=Batch)) + ###with jitter overplotted, symbol shape defined by batch
  geom_text(aes(x=Strain, y=-3, label=Letter),color="black", data=auc_cld) + ###Get the letters from auc_cld 
  #and write those to position y=-3
  labs(y="AUDPC") + #Y-Axis label
  ggtitle("b) AUDPC raw values and estimated means with 95% CI")+
  theme_bw() #Title 

###Arrage figures via Viewports
pushViewport(viewport(layout= grid.layout(nrow=1, ncol=2)))
print(Fig2a, vp = viewport(layout.pos.row = 1, layout.pos.col = 1))
print(Fig2b, vp = viewport(layout.pos.row = 1, layout.pos.col = 2))

###End of Fig2####
####clear device
dev.off()

#### Fig 3 ####
###Use only 6 Strains to make the facetting more usable, its kind of not nice when the number of plots is a prime..

#Omit Strain4, it is not super interesting, its basically strain3 +1
Fig3Strains <- c("Strain1","Strain2","Strain3","Strain5","Strain6","Strain7")

#predict uncensored
di_long <- add_predictions(di_long, lmer(DI ~ Strain + Strain:DPI + (1 | subject) + (1 | Batch), data = di_long), var="lmer.pred.no.cens")

###3A: Linear fits
F.3A <- ggplot(data=filter(di_long, Strain %in% Fig3Strains), aes(x=DPI,y=DI)) +
  geom_jitter(aes(group=DPI),data=filter(di_long, Strain %in% Fig3Strains, Useful=="No", Strain %in% Fig3Strains),color="red",alpha=0.3) +
  geom_jitter(aes(group=DPI),data=filter(di_long, Strain %in% Fig3Strains, Useful=="Yes", Strain %in% Fig3Strains),color="darkgrey",alpha=0.5)+
  stat_summary(aes(y=lmer.pred),color="black", fun.y=mean , geom = "line" , alpha=0.6) +
  stat_summary(aes(y=lmer.pred.no.cens), fun.y=mean,color="red" , geom = "line" , alpha=0.6) +
  coord_cartesian(ylim=c(-0.5,4.5)) +
  facet_wrap(~Strain)+
  labs(x = "Days post infection", y = "Disease Index") + #Labels
  ggtitle("b) Raw data and linear fit for individual strains averaged across batches")+
  theme_bw()

###3B

F.3B <- ggplot(filter(di_summary, Strain %in% Fig3Strains), ###Use Only Batch A (too busy otherwise)
  aes(x=DPI,y=mean,color=Strain)) + ###Color by Strain, specify x and y.
  stat_summary(aes(fill=Strain), position="identity",alpha=0.15,fun.y=mean,geom="area") + ###Area plot, colored by Strain
  ##geom_errorbar(aes(x=DPI, ymax=mean+se, ymin=mean-se), size=0.25)+ ###This line adds SE. 
  ##geom_errorbar(aes(x=DPI, ymax=mean+sd, ymin=mean-sd), size=0.25)+ ###This line adds SD. 
  stat_summary(aes(y=lmer.pred, x=DPI), color="black", filter(di_long, Strain %in% Fig3Strains), fun.y = mean, geom="line") +
  stat_summary(aes(y=lmer.pred.no.cens, x=DPI), color="red", filter(di_long, Strain %in% Fig3Strains), fun.y = mean, geom="line") +
  facet_wrap(~Strain) +###One plot per strain 
  coord_cartesian(ylim=c(-0.5,4.5)) +
  labs(x = "Days post infection", y = "Disease Index") + #Labels
  ggtitle("a) Disease Areas and linear fits per strain averaged across batches")+ #Title+
  theme_bw() +
  theme(legend.position="none")


pushViewport(viewport(layout= grid.layout(nrow=1, ncol=2)))
print(F.3B, vp = viewport(layout.pos.row = 1, layout.pos.col = 1))
print(F.3A, vp = viewport(layout.pos.row = 1, layout.pos.col = 2))

###End of Fig3####
####clear device
dev.off()

#### Fig 4 ####

##Display MCP + CI from AUDPC and DISEASE_LMER
##Calculations are done in the Rmd and available through the global environment

gg_PWCI_AUDPC <- ggplot(pairwise_confint_AUDPC_lmer, aes(x = Comparison, y = Estimate, ymin = lwr, ymax = upr, color = abs(Estimate))) + ###Plot Comparison on x, estimate on y
  scale_x_discrete(limits = rev(levels(as.factor(pairwise_confint_AUDPC_lmer$Comparison)))) + ###Rescale x, so the order is inverted 
  geom_errorbar() + geom_point() + ###Draw data
  coord_flip() +  xlab("") +###Invert X and Y, hide xlab
  geom_hline(yintercept=0,linetype=2) + ###Horizontal because the plot is flipped...
  ggtitle("a) Difference in estimated means of the AUDPC")+ ##Add a title+
  theme_bw()+
  theme(legend.position="none") #No legend.

gg_icep_CI <- ggplot(confint_icep, aes(x = Strain, y = Estimate, ymin = lwr, ymax = upr, color = abs(Estimate))) + ###Plot Comparison on x, estimate on y
  scale_x_discrete(limits = rev(levels(as.factor(confint_icep$Strain)))) + ###Rescale x, so the order is inverted 
  geom_errorbar() + geom_point() + ###Draw data
  coord_flip()  + xlab("") +###Invert X and Y, hide legend
  ggtitle("b) Intercepts with 95%CI") + theme_bw()+
  theme(legend.position="none") #No legend.

gg_slopes_CI <- ggplot(confint_slopes, aes(x = Strain, y = Estimate, ymin = lwr, ymax = upr, color = abs(Estimate))) + ###Plot Comparison on x, estimate on y
  scale_x_discrete(limits = rev(levels(as.factor(confint_slopes$Strain)))) + ###Rescale x, so the order is inverted 
  geom_errorbar() + geom_point() + ###Draw data
  coord_flip() +  theme(legend.position="none") + xlab("") +###Invert X and Y, hide legend
  ggtitle("c) Slopes with 95%CI") + theme_bw()+
  theme(legend.position="none") #No legend.

pushViewport(viewport(layout= grid.layout(nrow=2, ncol=2)))
print(gg_PWCI_AUDPC, vp = viewport(layout.pos.row = c(1,2), layout.pos.col = 1))
print(gg_icep_CI, vp = viewport(layout.pos.row = 1, layout.pos.col = 2))
print(gg_slopes_CI, vp = viewport(layout.pos.row = 2, layout.pos.col = 2))
###End of Fig4####
####clear device
dev.off()


###Fig 5 ####

##Should display effects of bad cutoffing


###Because i need to make multiple survfit dataframes, i have converted the script used in the other sections into a function, hopefully this makes it a little easier to follow. 
###Working with multistrata survfits is a little annoying, and breaking the strata back into the predictors is not part of this fuction.
###I would hope that at some point after this was written, a proper survfit tidier has been included into broom.
survfit_to_df <- function(x) {
  strata_dummy <-NULL
  for(i in 1:length(x$strata)){
    # add vector for one strata according to number of rows of strata
    strata_dummy <- c(strata_dummy, rep(names(x$strata)[i], x$strata[i]))
  }
  #make x.df from x..
  x.df <- data.frame( 
    time = x$time, 
    n.risk = x$n.risk, 
    n.event = x$n.event, 
    surv = x$surv, 
    strata = strata_dummy, 
    upper = x$upper, 
    lower = x$lower 
  ) 
  zeros <- data.frame(time = 0, surv = 1, strata = names((x$strata)), 
                      upper = 1, lower = 1)
  
  x.df <- plyr::rbind.fill(zeros, x.df) 
  rm(strata_dummy)
  rm(zeros)
  return(x.df)
}


####Generate intermediate survival tables#####
surv_table1 <- data.frame(Subject=disease_index$subject,
                          Strain=disease_index$Strain,
                          Plant=disease_index$Plant,
                          Batch=disease_index$Batch)

###Fill survival table based on the di_long table. This generates warnings. These can be ignored and come from the min()
cutoff <- c(1)
for (i in 1:max(di_long$subject)) { #Go by subject
  dummy <- di_long[di_long$subject==i,] #generate dummy for the subject
  if (is.infinite(min(dummy$DPI[which(dummy$DI >= cutoff)]))) { #If none of the DI is greater than the cutoff (this is where warnings are generated, min on an empty object returns infinite!)
    surv_table1$End[i] <- max(dummy$DPI) #Generate a  observation, censoring at the maximum DPI recorded
    surv_table1$Death[i] <- 0 #Still alive, because it did not pass the cutoff
  } else { #If more than zero DI are greater than the cutoff
    surv_table1$End[i] <- min(dummy$DPI[which(dummy$DI >= cutoff)]) #Use the lowest DPI where condition is met
    surv_table1$Death[i] <- 1 #record as dead
  }
}

surv_table2 <- data.frame(Subject=disease_index$subject,
                         Strain=disease_index$Strain,
                         Plant=disease_index$Plant,
                         Batch=disease_index$Batch)
###Fill survival table based on the di_long table. This generates warnings. These can be ignored and come from the min()
cutoff <- c(2)
for (i in 1:max(di_long$subject)) { #Go by subject
  dummy <- di_long[di_long$subject==i,] #generate dummy for the subject
  if (is.infinite(min(dummy$DPI[which(dummy$DI >= cutoff)]))) { #If none of the DI is greater than the cutoff (this is where warnings are generated, min on an empty object returns infinite!)
    surv_table2$End[i] <- max(dummy$DPI) #Generate a  observation, censoring at the maximum DPI recorded
    surv_table2$Death[i] <- 0 #Still alive, because it did not pass the cutoff
  } else { #If more than zero DI are greater than the cutoff
    surv_table2$End[i] <- min(dummy$DPI[which(dummy$DI >= cutoff)]) #Use the lowest DPI where condition is met
    surv_table2$Death[i] <- 1 #record as dead
  }
}


surv_table3 <- data.frame(Subject=disease_index$subject,
                          Strain=disease_index$Strain,
                          Plant=disease_index$Plant,
                          Batch=disease_index$Batch)
###Fill survival table based on the di_long table. This generates warnings. These can be ignored and come from the min()
cutoff <- c(3)
for (i in 1:max(di_long$subject)) { #Go by subject
  dummy <- di_long[di_long$subject==i,] #generate dummy for the subject
  if (is.infinite(min(dummy$DPI[which(dummy$DI >= cutoff)]))) { #If none of the DI is greater than the cutoff (this is where warnings are generated, min on an empty object returns infinite!)
    surv_table3$End[i] <- max(dummy$DPI) #Generate a  observation, censoring at the maximum DPI recorded
    surv_table3$Death[i] <- 0 #Still alive, because it did not pass the cutoff
  } else { #If more than zero DI are greater than the cutoff
    surv_table3$End[i] <- min(dummy$DPI[which(dummy$DI >= cutoff)]) #Use the lowest DPI where condition is met
    surv_table3$Death[i] <- 1 #record as dead
  }
}



surv_table4 <- data.frame(Subject=disease_index$subject,
                          Strain=disease_index$Strain,
                          Plant=disease_index$Plant,
                          Batch=disease_index$Batch)
###Fill survival table based on the di_long table. This generates warnings. These can be ignored and come from the min()
cutoff <- c(4)
for (i in 1:max(di_long$subject)) { #Go by subject
  dummy <- di_long[di_long$subject==i,] #generate dummy for the subject
  if (is.infinite(min(dummy$DPI[which(dummy$DI >= cutoff)]))) { #If none of the DI is greater than the cutoff (this is where warnings are generated, min on an empty object returns infinite!)
    surv_table4$End[i] <- max(dummy$DPI) #Generate a  observation, censoring at the maximum DPI recorded
    surv_table4$Death[i] <- 0 #Still alive, because it did not pass the cutoff
  } else { #If more than zero DI are greater than the cutoff
    surv_table4$End[i] <- min(dummy$DPI[which(dummy$DI >= cutoff)]) #Use the lowest DPI where condition is met
    surv_table4$Death[i] <- 1 #record as dead
  }
}

###Make survfits with the four survival tables..
surv_DI_fit1 <- with(surv_table1, survfit(Surv(End, Death) ~Strain+Batch))
surv_DI_fit2 <- with(surv_table2, survfit(Surv(End, Death) ~Strain+Batch))
surv_DI_fit3 <- with(surv_table3, survfit(Surv(End, Death) ~Strain+Batch))
surv_DI_fit4 <- with(surv_table4, survfit(Surv(End, Death) ~Strain+Batch))
###Make survfits into surv_df
surv_DI_fit1.df <- survfit_to_df(surv_DI_fit1)
surv_DI_fit2.df <- survfit_to_df(surv_DI_fit2)
surv_DI_fit3.df <- survfit_to_df(surv_DI_fit3)
surv_DI_fit4.df <- survfit_to_df(surv_DI_fit4)

###However the strata field still needs to be split manually....

surv_DI_fit1.df$Batch <- as.factor( str_split_fixed(
  matrix( nrow=length(surv_DI_fit1.df$strata),ncol=2, unlist(strsplit(as.character(surv_DI_fit1.df$strata),", ")), byrow=T )[,2],"=",2)[,2])
surv_DI_fit1.df$Strain <- as.factor( str_split_fixed(
  matrix( nrow=length(surv_DI_fit1.df$strata),ncol=2, unlist(strsplit(as.character(surv_DI_fit1.df$strata),", ")), byrow=T )[,1],"=",2)[,2])
surv_DI_fit2.df$Batch <- as.factor( str_split_fixed(
  matrix( nrow=length(surv_DI_fit2.df$strata),ncol=2, unlist(strsplit(as.character(surv_DI_fit2.df$strata),", ")), byrow=T )[,2],"=",2)[,2])
surv_DI_fit2.df$Strain <- as.factor( str_split_fixed(
  matrix( nrow=length(surv_DI_fit2.df$strata),ncol=2, unlist(strsplit(as.character(surv_DI_fit2.df$strata),", ")), byrow=T )[,1],"=",2)[,2])
surv_DI_fit3.df$Batch <- as.factor( str_split_fixed(
  matrix( nrow=length(surv_DI_fit3.df$strata),ncol=2, unlist(strsplit(as.character(surv_DI_fit3.df$strata),", ")), byrow=T )[,2],"=",2)[,2])
surv_DI_fit3.df$Strain <- as.factor( str_split_fixed(
  matrix( nrow=length(surv_DI_fit3.df$strata),ncol=2, unlist(strsplit(as.character(surv_DI_fit3.df$strata),", ")), byrow=T )[,1],"=",2)[,2])
surv_DI_fit4.df$Batch <- as.factor( str_split_fixed(
  matrix( nrow=length(surv_DI_fit4.df$strata),ncol=2, unlist(strsplit(as.character(surv_DI_fit4.df$strata),", ")), byrow=T )[,2],"=",2)[,2])
surv_DI_fit4.df$Strain <- as.factor( str_split_fixed(
  matrix( nrow=length(surv_DI_fit4.df$strata),ncol=2, unlist(strsplit(as.character(surv_DI_fit4.df$strata),", ")), byrow=T )[,1],"=",2)[,2])

###End of intermediate table generation####

####PlotZone####
Fig5a <- ggplot(data=filter(surv_DI_fit1.df, Batch=="A   ", Strain %in% c("Strain1","Strain3", "Strain4") ), aes(time, surv, colour = Strain)) + 
  geom_step(aes(y = surv*100),alpha=0.75) +
  ggtitle("a) Event defined as DI >=1") +
  theme_bw() +
  labs(x = "Days post infection", y = "Percentage of plants alive") + 
  theme(legend.position="none")

Fig5b <- ggplot(data=filter(surv_DI_fit2.df, Batch=="A   ", Strain %in% c("Strain1","Strain3", "Strain4") ), aes(time, surv, colour = Strain)) + 
  geom_step(aes(y = surv*100),alpha=0.75) +
  ggtitle("b) Event defined as DI >= 2") +
  theme_bw() +
  labs(x = "Days post infection", y = "Percentage of plants alive") + 
  theme(legend.position="none")

Fig5c <- ggplot(data=filter(surv_DI_fit3.df, Batch=="A   ", Strain %in% c("Strain1","Strain3", "Strain4") ), aes(time, surv, colour = Strain)) + 
  geom_step(aes(y = surv*100),alpha=0.75) +
  ggtitle("c) Event defined as DI >= 3") +
  theme_bw() +
  labs(x = "Days post infection", y = "Percentage of plants alive") + 
  theme(legend.position="none")

Fig5d <- ggplot(data=filter(surv_DI_fit4.df, Batch=="A   ", Strain %in% c("Strain1","Strain3", "Strain4") ), aes(time, surv, colour = Strain)) + 
  geom_step(aes(y = surv*100),alpha=0.75) +
  ggtitle("d) Event defined as DI >= 4") +
  theme_bw() +
  labs(x = "Days post infection", y = "Percentage of plants alive") + 
  theme(legend.position="none")

pushViewport(viewport(layout= grid.layout(nrow=2, ncol=2)))
print(Fig5a, vp = viewport(layout.pos.row = 1, layout.pos.col = 1))
print(Fig5b, vp = viewport(layout.pos.row = 1, layout.pos.col = 2))
print(Fig5c, vp = viewport(layout.pos.row = 2, layout.pos.col = 1))
print(Fig5d, vp = viewport(layout.pos.row = 2, layout.pos.col = 2))

dev.off()
###End of Fig5####
####clear device
dev.off()



###Fig 6 - Survival Regression####


###Here come the Survregs
###Dont plot "Strain7" for space reasons
surv_strains <- c("Strain1","Strain2","Strain3","Strain5","Strain6","Strain4")

kmdata <- filter(surv_DI_fit.df, Strain %in% surv_strains)
curvdata <- filter(sreg_curves, Strain %in% surv_strains)
#Labels

ggwei <- ggplot(kmdata, aes(time, surv, colour = Strain)) +
  geom_step(aes(linetype=Batch)) +
  geom_line(data=curvdata,aes(y=wei.sreg),color="black") +
  facet_wrap(~Strain) + theme(legend.position="none")+
  ggtitle("c) Kaplan-Meier estimates and fit to Weibull distribution")+
  theme_bw()+
  labs(x = "Days post infection", y = "Fraction of plants alive") + 
  theme(legend.position="none")
###Plot of KM+Logistic
gglogis <- ggplot(kmdata, aes(time, surv, colour = Strain)) +
  geom_step(aes(linetype=Batch)) +
  geom_line(data=curvdata,aes(y=logis.sreg),color="black") +
  facet_wrap(~Strain) + theme(legend.position="none")+
  ggtitle("d) Kaplan-Meier estimates and fit to Logistic distribution")+
  theme_bw()+
  labs(x = "Days post infection", y = "Fraction of plants alive") + 
  theme(legend.position="none")
###Plot of KM+Gaussian  
gggaus <- ggplot(kmdata, aes(time, surv, colour = Strain)) +
  geom_step(aes(linetype=Batch)) +
  geom_line(data=curvdata,aes(y=gaus.sreg),color="black") +
  facet_wrap(~Strain) + theme(legend.position="none")+
  ggtitle("b) Kaplan-Meier estimates and fit to Gaussian distribution")+
  theme_bw()+
  labs(x = "Days post infection", y = "Fraction of plants alive") + 
  theme(legend.position="none")
###Plot of KM+Lognormal  
gglnorm <- ggplot(kmdata, aes(time, surv, colour = Strain)) +
  geom_step(aes(linetype=Batch)) +
  geom_line(data=curvdata,aes(y=lnorm.sreg),color="black") +
  facet_wrap(~Strain) + theme(legend.position="none")+
  ggtitle("a) Kaplan-Meier estimates and fit to Lognormal distribution")+
  theme_bw()+
  labs(x = "Days post infection", y = "Fraction of plants alive") + 
  theme(legend.position="none")

pushViewport(viewport(layout= grid.layout(nrow=2, ncol=2)))
print(gglnorm, vp = viewport(layout.pos.row = 1, layout.pos.col = 1))
print(gggaus, vp = viewport(layout.pos.row = 1, layout.pos.col = 2))
print(ggwei, vp = viewport(layout.pos.row = 2, layout.pos.col = 1))
print(gglogis, vp = viewport(layout.pos.row = 2, layout.pos.col = 2))
###End of Figure6###
##Clear device.
dev.off()

####MOdified Figure 6
lnorm.di <- cbind(as.character(sregparams$Strain[1:7]), 
                  round(as.numeric(sregparams$scale.lnorm[1:7]), digits = 4), 
                  round(as.numeric(sregparams$shape.lnorm[1:7]),  digits = 4),
                  round( as.numeric(exp(sregparams$shape.lnorm[1:7])), digits = 4))
colnames(lnorm.di) <- c("Strain","Scale","Shape","exp(Shape)")

gglnorm <- ggplot(kmdata, aes(time, surv)) +
  geom_step(aes(linetype=Batch)) +
  geom_line(data=curvdata,aes(y=lnorm.sreg),color="black") +
  facet_wrap(~Strain) + theme(legend.position="none")+
  ggtitle("Kaplan-Meier estimates and fit to Lognormal distribution")+
  theme_bw()+
  labs(x = "Days post infection", y = "Fraction of plants alive") + 
  theme(legend.position="none")

library(gridExtra)

# Set theme to allow for plotmath expressions
tt <- ttheme_default(colhead=list(fg_params = list(parse=TRUE)))
tbl <- tableGrob(lnorm.di, rows=NULL, theme=tt)
# Plot chart and table into one object
grid.arrange(tbl, gglnorm, 
             ncol=2,
             as.table=TRUE,
             heights=c(3,1))