rNIX/R/hypoAfrica.R

#' Add binomial confidence intervals to var, fraction estimates, evaluated over nVar participants
#'
#' @param df a data frame that contains var and nVar columns
#' @param var variable to calculate CI for, should be a fraction 0..1
#' @param nVar variable that contains number of participants
#'
#' @return data frame with added var_low and var_high columns containing low and high CI95 values
#'
#' @export

add.conf.int<-function(df,var,nVar='total'){
   v<-df[,var]
   n<-df[,nVar]
   y1<-base::rep(0, base::length(v))
   y2<-base::rep(0, base::length(v))
   for (i in 1:base::length(v)){
      z<-stats::binom.test(base::as.integer(v[i]*n[i]),n[i],0.5,"two.sided",0.95)
      y1[i]<-z$conf.int[1]
      y2[i]<-z$conf.int[2]
   }
   
   df[,base::sprintf('%s_low',var)]=y1
   df[,base::sprintf('%s_high',var)]=y2
   df
}

#' Add values from reference (for statistical comparison)
#'
#' @param df data frame with a timeVar, target variable var and count variable nVar
#' @param dfRef data frame of same shape as df, considered reference for experiment
#' @param var a variable to set reference values for
#' @param timeVar variable containg ids of time points, where data will be merged
#' @param nVar variable containing counts
#'
#' @return data frame df with new coluns var_ref, var_nRef
#'
#' @export

add.Ref<-function(df,dfRef,var,timeVar='visitid',nVar='total'){
   pRef<-base::sprintf('%s_ref',var)
   df[,pRef]<-base::rep(0, base::nrow(df))
   df[,'nRef']<-base::rep(0, base::nrow(df))
   for (i in 1:base::nrow(df)){
      df[i,pRef]<-dfRef[dfRef[,timeVar]==df[i,timeVar],var]
      df[i,'nRef']<-dfRef[dfRef[,timeVar]==df[i,timeVar],nVar]
   }
   df
}

#' Add statistical significance to df/dfRef difference using binomial distribution
#'
#' @param df data frame with target variable var, timeVar and count variable nVar
#' @param dfRef a reference data frame of the same shape as df
#' @param var the target variable
#' @param timeVar variable indicating comparable visits 
#' @param nVar count variable
#'
#' @return data frame with var_p column of calculated signficance of difference (binom.test) and boolean var_s column notifying statistical significance at 0.05 level
#'
#' @export

add.p.values<-function(df,dfRef,var,timeVar='visitid',nVar='total'){
   df<-add.Ref(df,dfRef,var,timeVar,nVar)
   pName<-base::sprintf('%s_ref',var)
   pRef<-df[,pName]
   v<-df[,var]
   n<-df[,nVar]
   p<-base::rep(0, base::nrow(df))
   for (i in 1:base::nrow(df)){
      z<-stats::binom.test(base::as.integer(v[i]*n[i]),n[i],pRef[i],"two.sided",0.95)
      p[i]<-z$p.value
   }
   df[,base::sprintf('%s_p',var)]<-p
   df[,base::sprintf('%s_s',var)]<-df[,base::sprintf('%s_p',var)]<0.05
   df
}

#' Add statistical significance to df/dfRef difference using Fisher exact method
#'
#' @param df data frame with target variable var, timeVar and count variable nVar
#' @param dfRef a reference data frame of the same shape as df
#' @param var the target variable
#' @param timeVar variable indicating comparable visits 
#' @param nVar count variable
#'
#' @return data frame with var_pF column of calculated signficance of difference (fisher.test) and var_sF, coded (+<0.05,*<0.01,**<0.001) statistical significance 
#'
#' @export

add.p.values.fisher<-function(df,dfRef,var,timeVar='visitid',nVar='total'){
   
   df<-add.Ref(df,dfRef,var,timeVar,nVar)
   pRef<-df[,base::sprintf('%s_ref',var)]
   nRef<-df[,'nRef']
   v<-df[,var]
   n<-df[,nVar]
   p<-base::rep(0, base::nrow(df))
   for (i in 1:base::nrow(df)){
   	  dat <- base::data.frame(tox = base::c(pRef[i]*nRef[i],v[i]*n[i]),
  	  	noTox = base::c((1-pRef[i])*nRef[i], (1-v[i])*n[i]),
  	  	row.names = base::c("ref", "hypoAfrica"),
	  	stringsAsFactors = FALSE)
      z<-stats::fisher.test(dat)
      p[i]<-z$p.value
   }
   df[,base::sprintf('%s_pF',var)]<-p
   df[,base::sprintf('%s_sF',var)]<-''
   df[df[,base::sprintf('%s_pF',var)]<0.05,base::sprintf('%s_sF',var)]<-'(+)'
   df[df[,base::sprintf('%s_pF',var)]<0.01,base::sprintf('%s_sF',var)]<-'(*)'
   df[df[,base::sprintf('%s_pF',var)]<0.001,base::sprintf('%s_sF',var)]<-'(**)'
   df
}

#' Plot comparison of measured and reference fractions
#'
#' @param df data frame of measured fractions
#' @param var target fraction variable
#' @param timeVar variable measuring time
#' @param doseVar variable for selecting datasets
#' @param nVar variable where counts are stored
#' @param doseBreaks values to break plots to
#' @param doseLabels labels for legend for breaks
#' @param doseColors colors to use for different plots
#'
#' @return grob object with ggplot
#'
#' @export 
#'
#' @importFrom rlang .data
#'
get.ggplot<-function(df,var,timeVar='visitid',doseVar='dose',nVar='total',
   doseBreaks=base::c(59.9,60,62,100),doseLabels=base::c('ref','60 Gy','62 Gy','all'),doseColors=base::c('black','red','magenta','blue')){
   df<-add.conf.int(df,var,nVar)
   #dfRef<-add.conf.int(dfRef,var,nVar)

   ghigh<-base::sprintf("%s_high",var)
   glow<-base::sprintf("%s_low",var)
   ggplot2::ggplot()+
	#ggplot2::geom_ribbon(ggplot2::aes(x=.data[[timeVar]],ymin=.data[[glow]],ymax=.data[[ghigh]],dfRef,fill='grey70')+
	ggplot2::geom_ribbon(ggplot2::aes(x=.data[[timeVar]],ymin=.data[[glow]],ymax=.data[[ghigh]],fill=base::as.factor(.data[[doseVar]])),df,alpha=0.2)+
	#ggplot2::geom_line(ggplot2::aes(x=.data[[timeVar]],y=.data[[var]]),dfRef)+
	ggplot2::geom_line(ggplot2::aes(x=.data[[timeVar]],y=.data[[var]],color=base::as.factor(.data[[doseVar]])),df)+
	ggplot2::scale_color_manual(name = "Dose", breaks = doseBreaks, values = doseColors,labels=doseLabels)+
	ggplot2::scale_fill_manual(name = "Dose", breaks = doseBreaks, values = doseColors,labels=doseLabels)+
	ggplot2::theme(legend.position="bottom")+
   #ggplot2::xlab('Evaluation visit')+
   #ggplot2::scale_x_continuous(breaks=timeBreaks,limits=c(0,4),labels=timeLabels)+
   ggplot2::ylab('Portion with toxicity')
   
   
}

add.label<-function(df,var,labelVar){
   if (labelVar %in% colnames(df)){
      return (df);
   }
   df[,labelVar]=base::paste0(base::round(df[, var], digits = 2))
   df
}

#' Get table of values (as ggplot graphic object)
#'
#' @param df data frame with reference data
#' @param dfRef reference data frame 
#' @param var target variable to be shown
#' @param timeVar variable measuring time
#' @param doseVar variable for selecting datasets
#' @param nVar variable where counts are stored
#' @param doseBreaks which doses are in df
#' @param doseLabels how to label doses in legend
#' @param timeBreaks where time points are assigned
#' @param dfAdd additional data frame not checked for statistics (NULL ignored)
#'
#' @return grobj with values laid out as text
#'
#' @export

get.ggtext<-function(df,dfRef,var,timeVar='visitid',doseVar='dose',nVar='total', 
   doseBreaks=c(59.9,60,62,100),doseLabels=c('ref','60 Gy','62 Gy','all'),
   timeBreaks=c(0,1,3,4),dfAdd=NULL){

   df<-add.p.values.fisher(df,dfRef,var,timeVar,nVar)
   sf<-base::sprintf('%s_sF',var)
   dfRef[,sf]<-base::rep('',nrow(dfRef))

   labelVar=base::sprintf('%sLabel',var)
   df<-add.label(df,var,labelVar)
   dfRef<-add.label(dfRef,var,labelVar)

   colSel<-base::c(timeVar,labelVar,sf,doseVar)
   df1<-base::rbind(df[,colSel],dfRef[,colSel])

   if (!is.null(dfAdd)){
      dfAdd[,sf]<-base::rep('',nrow(dfAdd))
      dfAdd<-add.label(dfAdd,var,labelVar)
      df1<-base::rbind(df1[,colSel],dfAdd[,colSel])
   }
	ggplot2::ggplot()+
	ggplot2::geom_text(ggplot2::aes(x=.data[[timeVar]],y=base::as.factor(.data[[doseVar]]),
      label=base::paste0(.data[[labelVar]],.data[[sf]]),
      hjust=0.5, vjust=0.5),size=5,df1)+
	#ggplot2::scale_x_continuous(breaks=timeBreaks,limits=base::c(0,4))+
	ggplot2::scale_y_discrete(breaks=doseBreaks,labels=doseLabels)+
	ggplot2::theme_classic(base_size=20) +
	ggplot2::theme(axis.line = ggplot2::element_blank(),axis.ticks  = ggplot2::element_blank(),axis.title.y  = ggplot2::element_blank(), axis.title.x  = ggplot2::element_blank(),axis.text.x = ggplot2::element_text(color="white"))+
	ggplot2::theme(plot.title = ggplot2::element_text(size=11, hjust =-0.06))+
	ggplot2::theme(panel.background = ggplot2::element_rect(fill = "white"))
}