Nothing
R/Reporter-Inference.R
In docinfeR: Automatic Reporter for Inference Analysis
Defines functions
docinfeR
main
Documented in
docinfeR
library(officer)
library(nortest)
library(flextable)
library(broom)
library(tictoc)
tictoc::tic("Global")
# Inference Report
my.OS <- Sys.info()['sysname']
main <- function(path, var.type="all"){
#Read .csv File
if(path == ""){
path <- file.choose(new = FALSE)
}
file_in <- file(path,"r")
#file_out <- file(out.path <<- paste(path,"-REPORT-INFERENCE-",var.type,".docx",sep = ""),open="wt", blocking = FALSE, encoding ="UTF-8")
file_out <- file(paste(tempdir(),"docinfeR-REPORT-INFERENCE-",var.type,"-",format(Sys.time(), "%a-%b-%d-%X-%Y"),".docx",sep = ""),open="wt", blocking = FALSE, encoding ="UTF-8")
#defining the document name
if(grepl("linux",tolower(my.OS))){
my.time <- format(Sys.time(), "%Y-%m-%d_%H%M%S")
aux.file.path <- paste(path,"-REPORT-INFERENCE-",var.type,".docx",sep = "")
aux.file.report <- file_out
dir.create("/tmp/data/", showWarnings = FALSE)
aux.file.rdata <- paste("/tmp/data/", my.time,"-TEMP.DATA",sep="")
}else{
my.time <- format(Sys.time(), "%Y-%m-%d_%H%M%S")
aux.file.path <- paste(path,"-REPORT-INFERENCE-",var.type,".docx",sep = "")
dir.create("/tmp/data/", showWarnings = FALSE)
aux.file.rdata <- paste("/tmp/data/", my.time,"-TEMP.DATA",sep="")
}
my_doc <- officer::read_docx()
factor.factor <- function(vars.val.pass,var1type,var2type){
if(length(unique(vars.val.pass[,var1type]))==1 || length(unique(vars.val.pass[,var2type]))==1){
warning(Sys.time()," - var1:",paste(var1type)," - var2:",paste(var2type)," - [ERROR][INF] Error in stats::chisq.test: 'x' must at least have 2 elements\n\n")
return({0})
}else{
crosstbl <- table(vars.val.pass[,var1type],vars.val.pass[,var2type],dnn=c(var1type,var2type))
crosstbl_matrix <-as.matrix(crosstbl)
count_1<-which(crosstbl_matrix<5)
count_2<-which(crosstbl_matrix>=5)
ratio1 <- length(count_1)/(nrow(crosstbl_matrix)*ncol(crosstbl_matrix))
ratio2 <-length(count_2)/(nrow(crosstbl_matrix)*ncol(crosstbl_matrix))
#ATTENTION: Crosstable can get big!!
if(nrow(crosstbl_matrix)*ncol(crosstbl_matrix)<100){
val.test.inf <- flextable::flextable(as.data.frame.table(crosstbl_matrix))
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
officer::body_add_par(my_doc, paste("NOTE: TEST Crosstable of high dimension. Therefore, not available in this report!"),style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
#if there are zeros and 80% or less of the values are less than 5
#do Fisher exact test
if(any(crosstbl == 0 )){
if(ratio1>0.20 && ratio1 < 0.80){
val.test.inf <- flextable::flextable(broom::tidy(my.fisher <- stats::fisher.test(x = crosstbl, alternative = "two.sided",simulate.p.value = TRUE)))
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("The results of the analysis shows that Fisher test presents a p-value of ",round(my.fisher$p.value,3)," ",ifelse(my.fisher$p.value<0.05,paste("(p<0.05)."),paste("(p>0.05)."))," Thus, the null hipothesis is ",ifelse(my.fisher$p.value<0.05,paste("rejected"),paste("non-rejected"))," with a confidence level of 95%. We can conclude that both variables are ",ifelse(my.fisher$p.value<0.05,paste("dependent."),paste("independent.")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, paste("NOTE: TEST not robust due to high percentage of values inferior to 5 in crosstable: ",ratio1*100,"%\n\n",sep=""), style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else if(ratio1 < 0.20){
val.test.inf <- flextable::flextable(broom::tidy(my.fisher <- stats::fisher.test(x = crosstbl, alternative = "two.sided",simulate.p.value = TRUE)))
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("The results of the analysis shows that Fisher test presents a p-value of ",round(my.fisher$p.value,3)," ",ifelse(my.fisher$p.value<0.05,paste("(p<0.05)."),paste("(p>0.05)."))," Thus, the null hipothesis is ",ifelse(my.fisher$p.value<0.05,paste("rejected"),paste("non-rejected"))," with a confidence level of 95%. We can conclude that both variables are ",ifelse(my.fisher$p.value<0.05,paste("dependent."),paste("independent.")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
return({0})
}
}else{
if(ratio1>0.20 && ratio1 < 0.80){
val.test.inf <- flextable::flextable(broom::tidy(my.chisq <- stats::chisq.test(x = crosstbl)))####what happens if there are values in the crosstbl equal to 5
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("The results of the analysis shows that chi-squared test presents a p-value of ",round(my.chisq$p.value,3)," ",ifelse(my.chisq$p.value<0.05,paste("(p<0.05)."),paste("(p>0.05)."))," Thus, the null hipothesis is ",ifelse(my.chisq$p.value<0.05,paste("rejected"),paste("non-rejected"))," with a confidence level of 95%. We can conclude that both variables are ",ifelse(my.chisq$p.value<0.05,paste("dependent."),paste("independent.")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, paste("NOTE: TEST not robust due to high percentage of values inferior to 5 in crosstable: ",ratio1*100,"%\n\n",sep=""), style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else if(ratio1 < 0.20){
val.test.inf <- flextable::flextable(broom::tidy(my.chisq <- stats::chisq.test(x = crosstbl)))####what happens if there are values in the crosstbl equal to 5
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("The results of the analysis shows that chi-squared test presents a p-value of ",round(my.chisq$p.value,3)," ",ifelse(my.chisq$p.value<0.05,paste("(p<0.05)."),paste("(p>0.05)."))," Thus, the null hipothesis is ",ifelse(my.chisq$p.value<0.05,paste("rejected"),paste("non-rejected"))," with a confidence level of 95%. We can conclude that both variables are ",ifelse(my.chisq$p.value<0.05,paste("dependent."),paste("independent.")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
return({0})
}
}
}
}
factor.numeric.integer <- function(vars.val.pass,var1type,var2type){
#check normality for quantitative variable
#with Kolmogorov-Smirnov (lilliefors)
if(length(unique(stats::na.omit(vars.val.pass[,var2type])))<2 || length(stats::na.omit(vars.val.pass[,var2type]))<5){
return({0})
}else{
p.value <-nortest::lillie.test(stats::na.omit(vars.val.pass[,var2type]))$p.value###The K-S test is for a continuous distribution and so MYDATA
}
aux.data <- vars.val.pass[stats::complete.cases(vars.val.pass[,c(var1type,var2type)]),c(var1type,var2type)]
#should not contain any ties (repeated values).
#https://stats.stackexchange.com/questions/232011/ties-should-not-be-present-in-one-sample-kolmgorov-smirnov-test-in-r
if(p.value < 0.05){#not Normal
if(length(unique(aux.data[,var2type]))==1){
warning(Sys.time()," - var2:",paste(var2type)," - [ERROR][INF] Mann-Whitney-Wilcoxon or Kruskal cannot be applied to numeric vectors with only 1 unique value\n\n")
return({0})
}
if(length(unique(aux.data[,var1type]))==2){###What do you mean with unique(var1)==2?
#Mann-Whitney
test <- broom::tidy(my.wilcox <- stats::wilcox.test(aux.data[,var2type]~aux.data[,var1type])) #stats::wilcox.test(y~A) where y is numeric and A is A binary factor (independent)
test$data.name<-paste(var2type,"by",var1type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.wilcox$p.value,3),ifelse(round(my.wilcox$p.value,3)<0.05,paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups of "',var2type,'" are very different regarding "',var1type,'" variable.',sep=""),paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, the distributions are similar. Thus, it is possible to assume that groups of "',var2type,'" are similar regarding "',var1type,'" variable.',sep="")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else if(length(unique(aux.data[,var1type]))>2){
#Kruskal-Wallis
if(length(unique(aux.data[,var1type]))==1 || length(unique(aux.data[,var2type]))<=2){
warning(Sys.time()," - var1:",paste(var1type)," - [ERROR][INF] Error in stats::kruskal.test.default: all observations are in the same group\n\n")
return({0})
}else{
test <- broom::tidy(my.kruskal <- stats::kruskal.test(x=aux.data[,var2type],g=aux.data[,var1type]))#https://www.statmethods.net/stats/nonparametric.html #stats::kruskal.test(y~A) # where y1 is numeric and A is a factor
test$data.name<-paste(var2type,"by",var1type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.kruskal$p.value,3),ifelse(round(my.kruskal$p.value,3)<0.05,paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups are very different regarding "',var1type,'" variable.',sep=""),paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups have similar distribution regarding "',var1type,'" variable.',sep="")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}else{
return({0})
}
}else{#normal
vars.val.pass <- stats::na.omit(vars.val.pass)
if(length(unique(vars.val.pass[,var1type]))==2){
#ttest
tryCatch(stop(val.test.inf <- flextable::flextable(broom::tidy(my.t.test <- stats::t.test(vars.val.pass[,var2type]~vars.val.pass[,var1type], var.equal=TRUE)))), error = function(e) {warning(Sys.time()," - var1:",paste(var1type)," - var2:",paste(var2type)," - [ERROR][INF] ",e,"\n\n")}, finally = {
# stats::t.test(y~x) where y is numeric and x is a binary factor
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.t.test$p.value,3),' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups of "',var1type,'" have the same distribution regarding variable "',var2type,'".',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
})
}else{##What do you mean with unique(var1)>2?
#Anova
if(length(unique(vars.val.pass[,var1type]))<2){
warning(Sys.time()," - var1:",paste(var1type)," - [ERROR][INF] Error in contrasts<-: contrasts can be applied only to factors with 2 or more levels\n\n")
return({0})
}else{
test<- broom::tidy(my.aov <- stats::aov(vars.val.pass[,var2type] ~ vars.val.pass[,var1type])) #"Anova"
my.aov$p.value <- stats::summary.aov(my.aov)[[1]][["Pr(>F)"]][1]
if(is.null(my.aov$p.value)){
return({0})
}
test$call <- paste("stats::aov(formula = ",var2type," ~ ",var1type,")",sep = "")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.aov$p.value,3),ifelse(round(my.aov$p.value,3)>=0.05,paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups have the same distribution regarding "',var2type,'" variable.',sep=""),paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups do not have the same distribution regarding "',var2type,'" variable.',sep="")),sep = "")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}
}
}
numeric.integer.factor <- function(vars.val.pass,var1type,var2type){
#check normality for quantitative variable
#with Kolmogorov-Smirnov (lilliefors)
if(length(unique(stats::na.omit(vars.val.pass[,var1type])))<2 || length(stats::na.omit(vars.val.pass[,var1type]))<5){
return({0})
}else{
p.value <-nortest::lillie.test(stats::na.omit(vars.val.pass[,var1type]))$p.value###The K-S test is for a continuous distribution and so MYDATA
}
aux.data <- vars.val.pass[stats::complete.cases(vars.val.pass[,c(var1type,var2type)]),c(var1type,var2type)]
#should not contain any ties (repeated values).
#https://stats.stackexchange.com/questions/232011/ties-should-not-be-present-in-one-sample-kolmgorov-smirnov-test-in-r
if(p.value < 0.05){#not Normal
if(length(unique(aux.data[,var2type]))==2){
#Mann-Whitney
test <- broom::tidy(my.wilcox <- stats::wilcox.test(aux.data[,var1type]~aux.data[,var2type])) #stats::wilcox.test(y~A) where y is numeric and A is A binary factor (independent)
test$data.name<-paste(var1type,"by",var2type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.wilcox$p.value,3),ifelse(round(my.wilcox$p.value,3)<0.05,paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups of "',var2type,'" are very different regarding "',var1type,'" variable.',sep=""),paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, the distributions are similar. Thus, it is possible to assume that groups of "',var2type,'" are similar regarding "',var1type,'" variable.',sep="")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else if(length(unique(aux.data[,var2type]))>2){
#Kruskal-Wallis
if(length(unique(aux.data[,var2type]))==1 || length(unique(aux.data[,var1type]))==1){
warning(Sys.time()," - var2:",paste(var2type)," - [ERROR][INF] Error in stats::kruskal.test.default: all observations are in the same group\n\n")
return({0})
}else{
test <- broom::tidy(my.kruskal <- stats::kruskal.test(aux.data[,var1type]~aux.data[,var2type]))#https://www.statmethods.net/stats/nonparametric.html #stats::kruskal.test(y~A) # where y1 is numeric and A is a factor
test$data.name<-paste(var1type,"by",var2type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.kruskal$p.value,3),ifelse(round(my.kruskal$p.value,3)<0.05,paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups are very different regarding "',var1type,'" variable.',sep=""),paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups have similar distribution regarding "',var1type,'" variable.',sep="")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}else{
return({0})
}
}else{#normal
#take out NAs
vars.val.pass <- stats::na.omit(vars.val.pass)
if(length(unique(vars.val.pass[,var2type]))==2){###What do you mean with unique(var1)==2?
#ttest
tryCatch(stop(val.test.inf <- flextable::flextable(broom::tidy(my.t.test <- stats::t.test(vars.val.pass[,var1type]~vars.val.pass[,var2type], var.equal=TRUE)))), error = function(e) {warning(Sys.time()," - var1:",paste(var1type)," - var2:",paste(var2type)," - [ERROR][INF] ",e,"\n\n")}, finally = {
# stats::t.test(y~x) where y is numeric and x is a binary factor
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.t.test$p.value,3),' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups of "',var1type,'" have the same distribution regarding variable "',var2type,'".',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
})
}else if(length(unique(vars.val.pass[,var2type]))>2){##What do you mean with unique(var1)>2?
#Anova
if(length(unique(vars.val.pass[,var2type]))==1){
warning(Sys.time()," - ",paste(var2type)," - [ERROR][INF] Error in contrasts<-: contrasts can be applied only to factors with 2 or more levels\n\n")
return({0})
}else{
test<- broom::tidy(my.aov <- stats::aov(vars.val.pass[,var1type] ~ vars.val.pass[,var2type])) #"Anova"
my.aov$p.value <- stats::summary.aov(my.aov)[[1]][["Pr(>F)"]][1]
if(is.null(my.aov$p.value)){
return({0})
}
test$call <- paste("stats::aov(formula = ",var1type," ~ ",var2type,")",sep = "")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.aov$p.value,3),ifelse(round(my.aov$p.value,3)>=0.05,paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups have the same distribution regarding "',var2type,'" variable.',sep=""),paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups do not have the same distribution regarding "',var2type,'" variable.',sep="")),sep = "")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}else{
return({0})
}
}
}
numeric.integer.numeric.integer <- function(vars.val.pass,var1type,var2type){
vars.val.pass <- vars.val.pass[stats::complete.cases(vars.val.pass[,c(var1type,var2type)]),c(var1type,var2type)]
#check normality for quantitative variable
#with Kolmogorov-Smirnov (lilliefors)
if(length(unique(stats::na.omit(vars.val.pass[,var1type])))<2 || length(stats::na.omit(vars.val.pass[,var1type]))<5){
return({0})
}else{
p.value.v1 <-nortest::lillie.test(stats::na.omit(vars.val.pass[,var1type]))$p.value###The K-S test is for a continuous distribution and so MYDATA
}
#check normality for quantitative variable
#with Kolmogorov-Smirnov (lilliefors)
if(length(unique(stats::na.omit(vars.val.pass[,var2type])))<2 || length(stats::na.omit(vars.val.pass[,var2type]))<5){
return({0})
}else{
p.value.v2 <-nortest::lillie.test(stats::na.omit(vars.val.pass[,var2type]))$p.value###The K-S test is for a continuous distribution and so MYDATA
}
#check if the variables are continuous or discrete
if((class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "double") &&
(class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "double")){
if(p.value.v1>0.05 && p.value.v2>0.05){
#Pearson Correlation
test <- broom::tidy(my.pearson <- stats::cor.test(vars.val.pass[,var1type],vars.val.pass[,var2type],method = "pearson"))
test$call <- paste(var1type,"and",var2type,")",sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("To analyze two numerical variables, correlations analysis is applied. Since ",var1type ,'" and "',var2type,'" have normal distribution, Pearson correlation is more appropriate.',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("Pearson correlation presents a p-value of ",round(my.pearson$p.value,3),". This means ",ifelse(my.pearson$p.value<0.05,"a significant correlation between both variables exists."," that the independence of both variables exists."),". The value of the correlation is ",round(my.pearson$estimate,3),". Since the correlation result is ",ifelse(my.pearson$estimate>=0,"positive","negative"),", both variables are ",ifelse(my.pearson$estimate>=0,"positively","negatively")," correlated, i.e., when ",var1type," increases, ",var2type," ",ifelse(my.pearson$estimate>=0,"increases.","decreases."),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
#Spearman Correlation
test <- broom::tidy(my.Spearman <- stats::cor.test(vars.val.pass[,var1type],vars.val.pass[,var2type],method = "spearman"))
test$call <- paste(var1type,"and",var2type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("To analyze two numerical variables, correlations analysis is applied. Since ",'"',var1type ,'" and/or "',var2type,'" do not have normal distribution, Spearman correlation is more appropriate.',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("Spearman correlation presents a p-value of ",round(my.Spearman$p.value,3),". This means that ",ifelse(my.Spearman$p.value<0.05,"a significant correlation between both variables exists.","the variables are independent.")," The value of the correlation is ",round(my.Spearman$estimate,3),". Since the correlation result is ",ifelse(my.Spearman$estimate>=0,"positive","negative"),", both variables are ",ifelse(my.Spearman$estimate>=0,"positively","negatively")," correlated, i.e., when ",var1type," increases, ",var2type," ",ifelse(my.Spearman$estimate>=0,"increases","decreases"),".",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}else if(class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var2type]) == "integer"){
#Spearman Correlation (one integer variable)
test <- broom::tidy(my.Spearman <- stats::cor.test(vars.val.pass[,var1type],vars.val.pass[,var2type],method = "spearman"))
test$call <- paste(var1type,"and",var2type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('To analyze two numerical variables (in this case, one of them is an integer), correlations analysis is applied. Since we have an integer variable, Spearman correlation is more appropriate.',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("Spearman correlation presents a p-value of ",round(my.Spearman$p.value,3),". This means that ",ifelse(my.Spearman$p.value<0.05,"a significant correlation between both variables exists."," the independence of both variables exists.")," The value of the correlation is ",round(my.Spearman$estimate,3),". Since the correlation result is ",ifelse(my.Spearman$estimate>=0,"positive","negative"),", both variables are ",ifelse(my.Spearman$estimate>=0,"positively","negatively")," correlated, i.e., when ",var1type," increases, ",var2type," ",ifelse(my.Spearman$estimate>=0,"increases","decreases"),".",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
return({0})
}
}
optimization.cycles.1 <- function(vars.val.pass,var1type,var2type){
while(1){
if(class(vars.val.pass[,var1type])=="factor" && class(vars.val.pass[,var2type])=="factor"){
factor.factor(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var1type])== "factor" && (class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
factor.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var2type])== "factor" && (class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double")){
numeric.integer.factor(vars.val.pass,var1type,var2type)
break
}
if((class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double") &&
(class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
numeric.integer.numeric.integer(vars.val.pass,var1type,var2type)
break
}
}
}
optimization.cycles.2 <- function(vars.val.pass,var1type,var2type){
while(1){
if(class(vars.val.pass[,var1type])== "factor" && (class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
factor.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var2type])== "factor" && (class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double")){
numeric.integer.factor(vars.val.pass,var1type,var2type)
break
}
if((class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double") &&
(class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
numeric.integer.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var1type])=="factor" && class(vars.val.pass[,var2type])=="factor"){
factor.factor(vars.val.pass,var1type,var2type)
break
}
}
}
optimization.cycles.3 <- function(vars.val.pass,var1type,var2type){
while(1){
if((class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double") &&
(class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
numeric.integer.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var1type])=="factor" && class(vars.val.pass[,var2type])=="factor"){
factor.factor(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var1type])== "factor" && (class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
factor.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var2type])== "factor" && (class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double")){
numeric.integer.factor(vars.val.pass,var1type,var2type)
break
}
}
}
inf.writing <- function(what="inference",vars.name.pass=NULL,vars.val.pass=NULL,analysis=NULL){
if(what=="inference"){
if(analysis=="header1"){
val.header1 <- paste("Inference Analysis for All Variables Combination in Dataset:", sep = "")
officer::body_add_par(my_doc, val.header1, style = "heading 1",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('An assessment of the normality of data is a prerequisite for many statistical tests because normal data is an underlying assumption in parametric tests. The variables used to test the normality need to be numerical. Additionally, to numerically test the normality, the length of the sample should be taken into account. If the length is smaller than 50 records we use "Shapiro-Wilk test" or else we use "Kolmogorov-Smirnov test".',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("Thus, normality test was done for the numerical variables. The results of the tests used -- Shapiro-Wilk or Kolmogorov-Smirnov test - allows us concluding that, with a 95% of confidence, the null hypothesis is rejected if p<0.05/non-rejected if p>0.05, i.e., there is evidence/there is no evidence to reject the null hypothesis and it may not be/ be considered the existence of normality, respectively.",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("In this sense, the numeric variable has not/has normal distribution and, consequently, non-parametric/parametric tests should be used with that variable.",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("To compare two or more categorical variables, a cross-tabulation (also called the contingency table) is the most adequate option. However, to analyze the statistical differences, the chi-squared test or fisher test, for independence, should be applied to the crosstab.",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("To analyze two numerical variables, correlations analysis is applied. If variables have/have not normal distribution, Pearson / Spearman correlation is more appropriate, respectively.",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
if(analysis=="header2"){
val.header2 <- paste('Inference Analysis between Variable "', vars.name.pass[1], '" and "',vars.name.pass[2],'":', sep = "")
officer::body_add_par(my_doc, val.header2, style = "heading 2",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
if(analysis=="res"){
var1type <- vars.name.pass[1]
var2type <- vars.name.pass[2]
#Optimization considering variable type to increase efficiency
nfactor.count <- length(which(sapply(vars.val.pass, class)=="factor"))
nnumeric.count <- length(which(sapply(vars.val.pass, class) %in% c("numeric","integer","double")))
if(nfactor.count > nnumeric.count)
{
optimization.cycles.1(vars.val.pass,var1type,var2type)
}else if(nfactor.count == nnumeric.count){
optimization.cycles.2(vars.val.pass,var1type,var2type)
}else{
optimization.cycles.3(vars.val.pass,var1type,var2type)
}
}
}
}
file1 <- file_in
if(is.null(file1)){return(NULL)}
a<-readLines(path,n=2)
b<- nchar(gsub(";", "", a[1]))
c<- nchar(gsub(",", "", a[1]))
d<- nchar(gsub("\t", "", a[1]))
if(b<c&b<d){
#r <- utils::read.table(path,sep=";",stringsAsFactors = FALSE,nrows=1,na.strings = c("NA",""))
r <- utils::read.table(path,sep=";",stringsAsFactors = FALSE,nrows=1)
SEP=";"
}else if(c<b&c<d){
#r <- utils::read.table(path,sep=",",stringsAsFactors = FALSE,nrows=1,na.strings = c("NA",""))
r <- utils::read.table(path,sep=",",stringsAsFactors = FALSE,nrows=1)
SEP=","
}else{
#r <- utils::read.table(path,sep="\t",stringsAsFactors = FALSE,nrows=1,na.strings = c("NA",""))
r <- utils::read.table(path,sep="\t",stringsAsFactors = FALSE,nrows=1)
SEP="\t"
}
#r1<-utils::read.table(path,sep=SEP,stringsAsFactors = FALSE,nrows=1,skip = 1,na.strings = c("NA",""))
r1<-utils::read.table(path,sep=SEP,stringsAsFactors = FALSE,nrows=1,skip = 1)
aciertos <- 0
for(i in 1: ncol(r)){
if(class(r[1,i])==class(r1[1,i])){
aciertos <- aciertos+1
}
}
HEADER=T
if(aciertos==ncol(r)){
HEADER=F
}
data <- utils::read.table(path,sep=SEP,header=HEADER)
#Inference Analysis
#variables combinations
data.vars.vector <- names(data)
#type of selected vars
if(var.type!="all"){
data.vars.vector <- data.vars.vector[which(sapply(data,class)==var.type)]
if (length(data.vars.vector) < 2){
warning(paste("Not enough variables of type", var.type, "to proceed with Inference Reporting. Suggestion: use var.type input equal to all", sep=" "))
return()
}
}
var.comb <- t(utils::combn(data.vars.vector,2))
inf.writing(what = "inference",vars.name.pass=NULL,vars.val.pass=NULL,analysis="header1")
tictoc::tic("Analysis")
for(comb in 1:nrow(var.comb)){
inf.writing(what = "inference",vars.name.pass=var.comb[comb,],vars.val.pass=data[,var.comb[comb,]],analysis="header2")
aux.res.inf <- inf.writing(what = "inference",vars.name.pass=var.comb[comb,],vars.val.pass=data[,var.comb[comb,]],analysis="res")
if(class(aux.res.inf)=="numeric"){
if(aux.res.inf==0){
officer::body_add_par(my_doc, "Analysis not possible due to data constraints!", style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}
}
save.image(file = paste(aux.file.rdata))
warning(Sys.time(),"Passed Inference\n\n")
tictoc::toc()
tictoc::tic("Writing DOC")
print(my_doc, target = paste(aux.file.path))
tictoc::toc()
save.image(file = paste(aux.file.rdata))
warning(Sys.time(),"Passed Final Print\n\n")
tictoc::toc()
if(grepl("lin",tolower(my.OS))){
close(aux.file.report)
}
warning("Passed Final Print")
sink(type = "message")
sink()
#end and print
}
#' Inference Report in .docx file
#'
#' This R Package asks for a .csv file with data and returns a report (.docx) with Inference Report concerning all possible combinations of variables (i.e. columns).
#'
#' @import officer nortest flextable broom tictoc stats utils
#' @importFrom Rdpack reprompt
#'
#' @param path (Optional) A character vector with the path to data file. If empty character string (""), interface will appear to choose file.
#' @param var.type (Optional) The type of variables to perform analysis, with possible values: "all", "numeric", "integer", "double", "factor", "character".
#' @return The output
#' will be a document in the temp folder (tempdir()).
#' @examples
#' \donttest{
#' library(docinfeR)
#' data(iris)
#' dir = tempdir()
#' write.csv(iris,file=paste(dir,"iriscsvfile.csv",sep=""))
#' docinfeR(path=paste(dir,"iriscsvfile.csv",sep=""))
#' }
#' @references
#' \insertAllCited{}
#' @export
docinfeR <- function(path="", var.type="all"){
main(path,var.type=var.type)
}
Try the docinfeR package in your browser
Any scripts or data that you put into this service are public.
docinfeR documentation built on July 1, 2020, 6:02 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions docinfeR main
Documented in docinfeR
library(officer)
library(nortest)
library(flextable)
library(broom)
library(tictoc)
tictoc::tic("Global")
# Inference Report
my.OS <- Sys.info()['sysname']
main <- function(path, var.type="all"){
#Read .csv File
if(path == ""){
path <- file.choose(new = FALSE)
}
file_in <- file(path,"r")
#file_out <- file(out.path <<- paste(path,"-REPORT-INFERENCE-",var.type,".docx",sep = ""),open="wt", blocking = FALSE, encoding ="UTF-8")
file_out <- file(paste(tempdir(),"docinfeR-REPORT-INFERENCE-",var.type,"-",format(Sys.time(), "%a-%b-%d-%X-%Y"),".docx",sep = ""),open="wt", blocking = FALSE, encoding ="UTF-8")
#defining the document name
if(grepl("linux",tolower(my.OS))){
my.time <- format(Sys.time(), "%Y-%m-%d_%H%M%S")
aux.file.path <- paste(path,"-REPORT-INFERENCE-",var.type,".docx",sep = "")
aux.file.report <- file_out
dir.create("/tmp/data/", showWarnings = FALSE)
aux.file.rdata <- paste("/tmp/data/", my.time,"-TEMP.DATA",sep="")
}else{
my.time <- format(Sys.time(), "%Y-%m-%d_%H%M%S")
aux.file.path <- paste(path,"-REPORT-INFERENCE-",var.type,".docx",sep = "")
dir.create("/tmp/data/", showWarnings = FALSE)
aux.file.rdata <- paste("/tmp/data/", my.time,"-TEMP.DATA",sep="")
}
my_doc <- officer::read_docx()
factor.factor <- function(vars.val.pass,var1type,var2type){
if(length(unique(vars.val.pass[,var1type]))==1 || length(unique(vars.val.pass[,var2type]))==1){
warning(Sys.time()," - var1:",paste(var1type)," - var2:",paste(var2type)," - [ERROR][INF] Error in stats::chisq.test: 'x' must at least have 2 elements\n\n")
return({0})
}else{
crosstbl <- table(vars.val.pass[,var1type],vars.val.pass[,var2type],dnn=c(var1type,var2type))
crosstbl_matrix <-as.matrix(crosstbl)
count_1<-which(crosstbl_matrix<5)
count_2<-which(crosstbl_matrix>=5)
ratio1 <- length(count_1)/(nrow(crosstbl_matrix)*ncol(crosstbl_matrix))
ratio2 <-length(count_2)/(nrow(crosstbl_matrix)*ncol(crosstbl_matrix))
#ATTENTION: Crosstable can get big!!
if(nrow(crosstbl_matrix)*ncol(crosstbl_matrix)<100){
val.test.inf <- flextable::flextable(as.data.frame.table(crosstbl_matrix))
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
officer::body_add_par(my_doc, paste("NOTE: TEST Crosstable of high dimension. Therefore, not available in this report!"),style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
#if there are zeros and 80% or less of the values are less than 5
#do Fisher exact test
if(any(crosstbl == 0 )){
if(ratio1>0.20 && ratio1 < 0.80){
val.test.inf <- flextable::flextable(broom::tidy(my.fisher <- stats::fisher.test(x = crosstbl, alternative = "two.sided",simulate.p.value = TRUE)))
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("The results of the analysis shows that Fisher test presents a p-value of ",round(my.fisher$p.value,3)," ",ifelse(my.fisher$p.value<0.05,paste("(p<0.05)."),paste("(p>0.05)."))," Thus, the null hipothesis is ",ifelse(my.fisher$p.value<0.05,paste("rejected"),paste("non-rejected"))," with a confidence level of 95%. We can conclude that both variables are ",ifelse(my.fisher$p.value<0.05,paste("dependent."),paste("independent.")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, paste("NOTE: TEST not robust due to high percentage of values inferior to 5 in crosstable: ",ratio1*100,"%\n\n",sep=""), style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else if(ratio1 < 0.20){
val.test.inf <- flextable::flextable(broom::tidy(my.fisher <- stats::fisher.test(x = crosstbl, alternative = "two.sided",simulate.p.value = TRUE)))
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("The results of the analysis shows that Fisher test presents a p-value of ",round(my.fisher$p.value,3)," ",ifelse(my.fisher$p.value<0.05,paste("(p<0.05)."),paste("(p>0.05)."))," Thus, the null hipothesis is ",ifelse(my.fisher$p.value<0.05,paste("rejected"),paste("non-rejected"))," with a confidence level of 95%. We can conclude that both variables are ",ifelse(my.fisher$p.value<0.05,paste("dependent."),paste("independent.")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
return({0})
}
}else{
if(ratio1>0.20 && ratio1 < 0.80){
val.test.inf <- flextable::flextable(broom::tidy(my.chisq <- stats::chisq.test(x = crosstbl)))####what happens if there are values in the crosstbl equal to 5
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("The results of the analysis shows that chi-squared test presents a p-value of ",round(my.chisq$p.value,3)," ",ifelse(my.chisq$p.value<0.05,paste("(p<0.05)."),paste("(p>0.05)."))," Thus, the null hipothesis is ",ifelse(my.chisq$p.value<0.05,paste("rejected"),paste("non-rejected"))," with a confidence level of 95%. We can conclude that both variables are ",ifelse(my.chisq$p.value<0.05,paste("dependent."),paste("independent.")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, paste("NOTE: TEST not robust due to high percentage of values inferior to 5 in crosstable: ",ratio1*100,"%\n\n",sep=""), style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else if(ratio1 < 0.20){
val.test.inf <- flextable::flextable(broom::tidy(my.chisq <- stats::chisq.test(x = crosstbl)))####what happens if there are values in the crosstbl equal to 5
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("The results of the analysis shows that chi-squared test presents a p-value of ",round(my.chisq$p.value,3)," ",ifelse(my.chisq$p.value<0.05,paste("(p<0.05)."),paste("(p>0.05)."))," Thus, the null hipothesis is ",ifelse(my.chisq$p.value<0.05,paste("rejected"),paste("non-rejected"))," with a confidence level of 95%. We can conclude that both variables are ",ifelse(my.chisq$p.value<0.05,paste("dependent."),paste("independent.")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
return({0})
}
}
}
}
factor.numeric.integer <- function(vars.val.pass,var1type,var2type){
#check normality for quantitative variable
#with Kolmogorov-Smirnov (lilliefors)
if(length(unique(stats::na.omit(vars.val.pass[,var2type])))<2 || length(stats::na.omit(vars.val.pass[,var2type]))<5){
return({0})
}else{
p.value <-nortest::lillie.test(stats::na.omit(vars.val.pass[,var2type]))$p.value###The K-S test is for a continuous distribution and so MYDATA
}
aux.data <- vars.val.pass[stats::complete.cases(vars.val.pass[,c(var1type,var2type)]),c(var1type,var2type)]
#should not contain any ties (repeated values).
#https://stats.stackexchange.com/questions/232011/ties-should-not-be-present-in-one-sample-kolmgorov-smirnov-test-in-r
if(p.value < 0.05){#not Normal
if(length(unique(aux.data[,var2type]))==1){
warning(Sys.time()," - var2:",paste(var2type)," - [ERROR][INF] Mann-Whitney-Wilcoxon or Kruskal cannot be applied to numeric vectors with only 1 unique value\n\n")
return({0})
}
if(length(unique(aux.data[,var1type]))==2){###What do you mean with unique(var1)==2?
#Mann-Whitney
test <- broom::tidy(my.wilcox <- stats::wilcox.test(aux.data[,var2type]~aux.data[,var1type])) #stats::wilcox.test(y~A) where y is numeric and A is A binary factor (independent)
test$data.name<-paste(var2type,"by",var1type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.wilcox$p.value,3),ifelse(round(my.wilcox$p.value,3)<0.05,paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups of "',var2type,'" are very different regarding "',var1type,'" variable.',sep=""),paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, the distributions are similar. Thus, it is possible to assume that groups of "',var2type,'" are similar regarding "',var1type,'" variable.',sep="")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else if(length(unique(aux.data[,var1type]))>2){
#Kruskal-Wallis
if(length(unique(aux.data[,var1type]))==1 || length(unique(aux.data[,var2type]))<=2){
warning(Sys.time()," - var1:",paste(var1type)," - [ERROR][INF] Error in stats::kruskal.test.default: all observations are in the same group\n\n")
return({0})
}else{
test <- broom::tidy(my.kruskal <- stats::kruskal.test(x=aux.data[,var2type],g=aux.data[,var1type]))#https://www.statmethods.net/stats/nonparametric.html #stats::kruskal.test(y~A) # where y1 is numeric and A is a factor
test$data.name<-paste(var2type,"by",var1type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.kruskal$p.value,3),ifelse(round(my.kruskal$p.value,3)<0.05,paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups are very different regarding "',var1type,'" variable.',sep=""),paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups have similar distribution regarding "',var1type,'" variable.',sep="")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}else{
return({0})
}
}else{#normal
vars.val.pass <- stats::na.omit(vars.val.pass)
if(length(unique(vars.val.pass[,var1type]))==2){
#ttest
tryCatch(stop(val.test.inf <- flextable::flextable(broom::tidy(my.t.test <- stats::t.test(vars.val.pass[,var2type]~vars.val.pass[,var1type], var.equal=TRUE)))), error = function(e) {warning(Sys.time()," - var1:",paste(var1type)," - var2:",paste(var2type)," - [ERROR][INF] ",e,"\n\n")}, finally = {
# stats::t.test(y~x) where y is numeric and x is a binary factor
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.t.test$p.value,3),' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups of "',var1type,'" have the same distribution regarding variable "',var2type,'".',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
})
}else{##What do you mean with unique(var1)>2?
#Anova
if(length(unique(vars.val.pass[,var1type]))<2){
warning(Sys.time()," - var1:",paste(var1type)," - [ERROR][INF] Error in contrasts<-: contrasts can be applied only to factors with 2 or more levels\n\n")
return({0})
}else{
test<- broom::tidy(my.aov <- stats::aov(vars.val.pass[,var2type] ~ vars.val.pass[,var1type])) #"Anova"
my.aov$p.value <- stats::summary.aov(my.aov)[[1]][["Pr(>F)"]][1]
if(is.null(my.aov$p.value)){
return({0})
}
test$call <- paste("stats::aov(formula = ",var2type," ~ ",var1type,")",sep = "")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.aov$p.value,3),ifelse(round(my.aov$p.value,3)>=0.05,paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups have the same distribution regarding "',var2type,'" variable.',sep=""),paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups do not have the same distribution regarding "',var2type,'" variable.',sep="")),sep = "")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}
}
}
numeric.integer.factor <- function(vars.val.pass,var1type,var2type){
#check normality for quantitative variable
#with Kolmogorov-Smirnov (lilliefors)
if(length(unique(stats::na.omit(vars.val.pass[,var1type])))<2 || length(stats::na.omit(vars.val.pass[,var1type]))<5){
return({0})
}else{
p.value <-nortest::lillie.test(stats::na.omit(vars.val.pass[,var1type]))$p.value###The K-S test is for a continuous distribution and so MYDATA
}
aux.data <- vars.val.pass[stats::complete.cases(vars.val.pass[,c(var1type,var2type)]),c(var1type,var2type)]
#should not contain any ties (repeated values).
#https://stats.stackexchange.com/questions/232011/ties-should-not-be-present-in-one-sample-kolmgorov-smirnov-test-in-r
if(p.value < 0.05){#not Normal
if(length(unique(aux.data[,var2type]))==2){
#Mann-Whitney
test <- broom::tidy(my.wilcox <- stats::wilcox.test(aux.data[,var1type]~aux.data[,var2type])) #stats::wilcox.test(y~A) where y is numeric and A is A binary factor (independent)
test$data.name<-paste(var1type,"by",var2type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.wilcox$p.value,3),ifelse(round(my.wilcox$p.value,3)<0.05,paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups of "',var2type,'" are very different regarding "',var1type,'" variable.',sep=""),paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, the distributions are similar. Thus, it is possible to assume that groups of "',var2type,'" are similar regarding "',var1type,'" variable.',sep="")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else if(length(unique(aux.data[,var2type]))>2){
#Kruskal-Wallis
if(length(unique(aux.data[,var2type]))==1 || length(unique(aux.data[,var1type]))==1){
warning(Sys.time()," - var2:",paste(var2type)," - [ERROR][INF] Error in stats::kruskal.test.default: all observations are in the same group\n\n")
return({0})
}else{
test <- broom::tidy(my.kruskal <- stats::kruskal.test(aux.data[,var1type]~aux.data[,var2type]))#https://www.statmethods.net/stats/nonparametric.html #stats::kruskal.test(y~A) # where y1 is numeric and A is a factor
test$data.name<-paste(var1type,"by",var2type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.kruskal$p.value,3),ifelse(round(my.kruskal$p.value,3)<0.05,paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups are very different regarding "',var1type,'" variable.',sep=""),paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups have similar distribution regarding "',var1type,'" variable.',sep="")),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}else{
return({0})
}
}else{#normal
#take out NAs
vars.val.pass <- stats::na.omit(vars.val.pass)
if(length(unique(vars.val.pass[,var2type]))==2){###What do you mean with unique(var1)==2?
#ttest
tryCatch(stop(val.test.inf <- flextable::flextable(broom::tidy(my.t.test <- stats::t.test(vars.val.pass[,var1type]~vars.val.pass[,var2type], var.equal=TRUE)))), error = function(e) {warning(Sys.time()," - var1:",paste(var1type)," - var2:",paste(var2type)," - [ERROR][INF] ",e,"\n\n")}, finally = {
# stats::t.test(y~x) where y is numeric and x is a binary factor
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.t.test$p.value,3),' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups of "',var1type,'" have the same distribution regarding variable "',var2type,'".',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
})
}else if(length(unique(vars.val.pass[,var2type]))>2){##What do you mean with unique(var1)>2?
#Anova
if(length(unique(vars.val.pass[,var2type]))==1){
warning(Sys.time()," - ",paste(var2type)," - [ERROR][INF] Error in contrasts<-: contrasts can be applied only to factors with 2 or more levels\n\n")
return({0})
}else{
test<- broom::tidy(my.aov <- stats::aov(vars.val.pass[,var1type] ~ vars.val.pass[,var2type])) #"Anova"
my.aov$p.value <- stats::summary.aov(my.aov)[[1]][["Pr(>F)"]][1]
if(is.null(my.aov$p.value)){
return({0})
}
test$call <- paste("stats::aov(formula = ",var1type," ~ ",var2type,")",sep = "")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('Analyses show that p-value is ',round(my.aov$p.value,3),ifelse(round(my.aov$p.value,3)>=0.05,paste(' (p>0.05). This means that null hypothesis is non-rejected (p>0.05) and, therefore, there are no statistical differences. Thus, it is possible to assume that groups have the same distribution regarding "',var2type,'" variable.',sep=""),paste(' (p<0.05). This means that null hypothesis is rejected (p<0.05) and, therefore, there are statistical differences. Thus, it is possible to assume that groups do not have the same distribution regarding "',var2type,'" variable.',sep="")),sep = "")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}else{
return({0})
}
}
}
numeric.integer.numeric.integer <- function(vars.val.pass,var1type,var2type){
vars.val.pass <- vars.val.pass[stats::complete.cases(vars.val.pass[,c(var1type,var2type)]),c(var1type,var2type)]
#check normality for quantitative variable
#with Kolmogorov-Smirnov (lilliefors)
if(length(unique(stats::na.omit(vars.val.pass[,var1type])))<2 || length(stats::na.omit(vars.val.pass[,var1type]))<5){
return({0})
}else{
p.value.v1 <-nortest::lillie.test(stats::na.omit(vars.val.pass[,var1type]))$p.value###The K-S test is for a continuous distribution and so MYDATA
}
#check normality for quantitative variable
#with Kolmogorov-Smirnov (lilliefors)
if(length(unique(stats::na.omit(vars.val.pass[,var2type])))<2 || length(stats::na.omit(vars.val.pass[,var2type]))<5){
return({0})
}else{
p.value.v2 <-nortest::lillie.test(stats::na.omit(vars.val.pass[,var2type]))$p.value###The K-S test is for a continuous distribution and so MYDATA
}
#check if the variables are continuous or discrete
if((class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "double") &&
(class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "double")){
if(p.value.v1>0.05 && p.value.v2>0.05){
#Pearson Correlation
test <- broom::tidy(my.pearson <- stats::cor.test(vars.val.pass[,var1type],vars.val.pass[,var2type],method = "pearson"))
test$call <- paste(var1type,"and",var2type,")",sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("To analyze two numerical variables, correlations analysis is applied. Since ",var1type ,'" and "',var2type,'" have normal distribution, Pearson correlation is more appropriate.',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("Pearson correlation presents a p-value of ",round(my.pearson$p.value,3),". This means ",ifelse(my.pearson$p.value<0.05,"a significant correlation between both variables exists."," that the independence of both variables exists."),". The value of the correlation is ",round(my.pearson$estimate,3),". Since the correlation result is ",ifelse(my.pearson$estimate>=0,"positive","negative"),", both variables are ",ifelse(my.pearson$estimate>=0,"positively","negatively")," correlated, i.e., when ",var1type," increases, ",var2type," ",ifelse(my.pearson$estimate>=0,"increases.","decreases."),sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
#Spearman Correlation
test <- broom::tidy(my.Spearman <- stats::cor.test(vars.val.pass[,var1type],vars.val.pass[,var2type],method = "spearman"))
test$call <- paste(var1type,"and",var2type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("To analyze two numerical variables, correlations analysis is applied. Since ",'"',var1type ,'" and/or "',var2type,'" do not have normal distribution, Spearman correlation is more appropriate.',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("Spearman correlation presents a p-value of ",round(my.Spearman$p.value,3),". This means that ",ifelse(my.Spearman$p.value<0.05,"a significant correlation between both variables exists.","the variables are independent.")," The value of the correlation is ",round(my.Spearman$estimate,3),". Since the correlation result is ",ifelse(my.Spearman$estimate>=0,"positive","negative"),", both variables are ",ifelse(my.Spearman$estimate>=0,"positively","negatively")," correlated, i.e., when ",var1type," increases, ",var2type," ",ifelse(my.Spearman$estimate>=0,"increases","decreases"),".",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}else if(class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var2type]) == "integer"){
#Spearman Correlation (one integer variable)
test <- broom::tidy(my.Spearman <- stats::cor.test(vars.val.pass[,var1type],vars.val.pass[,var2type],method = "spearman"))
test$call <- paste(var1type,"and",var2type,sep = " ")
val.test.inf <- flextable::flextable(test)
flextable::body_add_flextable(my_doc, flextable::autofit(flextable::fontsize(val.test.inf, size = 6, part = "all")), pos="after",split = TRUE)
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('To analyze two numerical variables (in this case, one of them is an integer), correlations analysis is applied. Since we have an integer variable, Spearman correlation is more appropriate.',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("Spearman correlation presents a p-value of ",round(my.Spearman$p.value,3),". This means that ",ifelse(my.Spearman$p.value<0.05,"a significant correlation between both variables exists."," the independence of both variables exists.")," The value of the correlation is ",round(my.Spearman$estimate,3),". Since the correlation result is ",ifelse(my.Spearman$estimate>=0,"positive","negative"),", both variables are ",ifelse(my.Spearman$estimate>=0,"positively","negatively")," correlated, i.e., when ",var1type," increases, ",var2type," ",ifelse(my.Spearman$estimate>=0,"increases","decreases"),".",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after") # blank paragraph
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}else{
return({0})
}
}
optimization.cycles.1 <- function(vars.val.pass,var1type,var2type){
while(1){
if(class(vars.val.pass[,var1type])=="factor" && class(vars.val.pass[,var2type])=="factor"){
factor.factor(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var1type])== "factor" && (class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
factor.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var2type])== "factor" && (class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double")){
numeric.integer.factor(vars.val.pass,var1type,var2type)
break
}
if((class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double") &&
(class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
numeric.integer.numeric.integer(vars.val.pass,var1type,var2type)
break
}
}
}
optimization.cycles.2 <- function(vars.val.pass,var1type,var2type){
while(1){
if(class(vars.val.pass[,var1type])== "factor" && (class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
factor.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var2type])== "factor" && (class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double")){
numeric.integer.factor(vars.val.pass,var1type,var2type)
break
}
if((class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double") &&
(class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
numeric.integer.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var1type])=="factor" && class(vars.val.pass[,var2type])=="factor"){
factor.factor(vars.val.pass,var1type,var2type)
break
}
}
}
optimization.cycles.3 <- function(vars.val.pass,var1type,var2type){
while(1){
if((class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double") &&
(class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
numeric.integer.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var1type])=="factor" && class(vars.val.pass[,var2type])=="factor"){
factor.factor(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var1type])== "factor" && (class(vars.val.pass[,var2type]) == "numeric"|| class(vars.val.pass[,var2type]) == "integer" || class(vars.val.pass[,var2type]) == "double")){
factor.numeric.integer(vars.val.pass,var1type,var2type)
break
}
if(class(vars.val.pass[,var2type])== "factor" && (class(vars.val.pass[,var1type]) == "numeric"|| class(vars.val.pass[,var1type]) == "integer" || class(vars.val.pass[,var1type]) == "double")){
numeric.integer.factor(vars.val.pass,var1type,var2type)
break
}
}
}
inf.writing <- function(what="inference",vars.name.pass=NULL,vars.val.pass=NULL,analysis=NULL){
if(what=="inference"){
if(analysis=="header1"){
val.header1 <- paste("Inference Analysis for All Variables Combination in Dataset:", sep = "")
officer::body_add_par(my_doc, val.header1, style = "heading 1",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste('An assessment of the normality of data is a prerequisite for many statistical tests because normal data is an underlying assumption in parametric tests. The variables used to test the normality need to be numerical. Additionally, to numerically test the normality, the length of the sample should be taken into account. If the length is smaller than 50 records we use "Shapiro-Wilk test" or else we use "Kolmogorov-Smirnov test".',sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("Thus, normality test was done for the numerical variables. The results of the tests used -- Shapiro-Wilk or Kolmogorov-Smirnov test - allows us concluding that, with a 95% of confidence, the null hypothesis is rejected if p<0.05/non-rejected if p>0.05, i.e., there is evidence/there is no evidence to reject the null hypothesis and it may not be/ be considered the existence of normality, respectively.",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("In this sense, the numeric variable has not/has normal distribution and, consequently, non-parametric/parametric tests should be used with that variable.",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("To compare two or more categorical variables, a cross-tabulation (also called the contingency table) is the most adequate option. However, to analyze the statistical differences, the chi-squared test or fisher test, for independence, should be applied to the crosstab.",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
aux.text.inf <- paste("To analyze two numerical variables, correlations analysis is applied. If variables have/have not normal distribution, Pearson / Spearman correlation is more appropriate, respectively.",sep="")
officer::body_add_par(my_doc, aux.text.inf, style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
if(analysis=="header2"){
val.header2 <- paste('Inference Analysis between Variable "', vars.name.pass[1], '" and "',vars.name.pass[2],'":', sep = "")
officer::body_add_par(my_doc, val.header2, style = "heading 2",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
if(analysis=="res"){
var1type <- vars.name.pass[1]
var2type <- vars.name.pass[2]
#Optimization considering variable type to increase efficiency
nfactor.count <- length(which(sapply(vars.val.pass, class)=="factor"))
nnumeric.count <- length(which(sapply(vars.val.pass, class) %in% c("numeric","integer","double")))
if(nfactor.count > nnumeric.count)
{
optimization.cycles.1(vars.val.pass,var1type,var2type)
}else if(nfactor.count == nnumeric.count){
optimization.cycles.2(vars.val.pass,var1type,var2type)
}else{
optimization.cycles.3(vars.val.pass,var1type,var2type)
}
}
}
}
file1 <- file_in
if(is.null(file1)){return(NULL)}
a<-readLines(path,n=2)
b<- nchar(gsub(";", "", a[1]))
c<- nchar(gsub(",", "", a[1]))
d<- nchar(gsub("\t", "", a[1]))
if(b<c&b<d){
#r <- utils::read.table(path,sep=";",stringsAsFactors = FALSE,nrows=1,na.strings = c("NA",""))
r <- utils::read.table(path,sep=";",stringsAsFactors = FALSE,nrows=1)
SEP=";"
}else if(c<b&c<d){
#r <- utils::read.table(path,sep=",",stringsAsFactors = FALSE,nrows=1,na.strings = c("NA",""))
r <- utils::read.table(path,sep=",",stringsAsFactors = FALSE,nrows=1)
SEP=","
}else{
#r <- utils::read.table(path,sep="\t",stringsAsFactors = FALSE,nrows=1,na.strings = c("NA",""))
r <- utils::read.table(path,sep="\t",stringsAsFactors = FALSE,nrows=1)
SEP="\t"
}
#r1<-utils::read.table(path,sep=SEP,stringsAsFactors = FALSE,nrows=1,skip = 1,na.strings = c("NA",""))
r1<-utils::read.table(path,sep=SEP,stringsAsFactors = FALSE,nrows=1,skip = 1)
aciertos <- 0
for(i in 1: ncol(r)){
if(class(r[1,i])==class(r1[1,i])){
aciertos <- aciertos+1
}
}
HEADER=T
if(aciertos==ncol(r)){
HEADER=F
}
data <- utils::read.table(path,sep=SEP,header=HEADER)
#Inference Analysis
#variables combinations
data.vars.vector <- names(data)
#type of selected vars
if(var.type!="all"){
data.vars.vector <- data.vars.vector[which(sapply(data,class)==var.type)]
if (length(data.vars.vector) < 2){
warning(paste("Not enough variables of type", var.type, "to proceed with Inference Reporting. Suggestion: use var.type input equal to all", sep=" "))
return()
}
}
var.comb <- t(utils::combn(data.vars.vector,2))
inf.writing(what = "inference",vars.name.pass=NULL,vars.val.pass=NULL,analysis="header1")
tictoc::tic("Analysis")
for(comb in 1:nrow(var.comb)){
inf.writing(what = "inference",vars.name.pass=var.comb[comb,],vars.val.pass=data[,var.comb[comb,]],analysis="header2")
aux.res.inf <- inf.writing(what = "inference",vars.name.pass=var.comb[comb,],vars.val.pass=data[,var.comb[comb,]],analysis="res")
if(class(aux.res.inf)=="numeric"){
if(aux.res.inf==0){
officer::body_add_par(my_doc, "Analysis not possible due to data constraints!", style = "Normal",pos="after")
officer::body_add_par(my_doc, "", style = "Normal",pos="after") # blank paragraph
}
}
}
save.image(file = paste(aux.file.rdata))
warning(Sys.time(),"Passed Inference\n\n")
tictoc::toc()
tictoc::tic("Writing DOC")
print(my_doc, target = paste(aux.file.path))
tictoc::toc()
save.image(file = paste(aux.file.rdata))
warning(Sys.time(),"Passed Final Print\n\n")
tictoc::toc()
if(grepl("lin",tolower(my.OS))){
close(aux.file.report)
}
warning("Passed Final Print")
sink(type = "message")
sink()
#end and print
}
#' Inference Report in .docx file
#'
#' This R Package asks for a .csv file with data and returns a report (.docx) with Inference Report concerning all possible combinations of variables (i.e. columns).
#'
#' @import officer nortest flextable broom tictoc stats utils
#' @importFrom Rdpack reprompt
#'
#' @param path (Optional) A character vector with the path to data file. If empty character string (""), interface will appear to choose file.
#' @param var.type (Optional) The type of variables to perform analysis, with possible values: "all", "numeric", "integer", "double", "factor", "character".
#' @return The output
#' will be a document in the temp folder (tempdir()).
#' @examples
#' \donttest{
#' library(docinfeR)
#' data(iris)
#' dir = tempdir()
#' write.csv(iris,file=paste(dir,"iriscsvfile.csv",sep=""))
#' docinfeR(path=paste(dir,"iriscsvfile.csv",sep=""))
#' }
#' @references
#' \insertAllCited{}
#' @export
docinfeR <- function(path="", var.type="all"){
main(path,var.type=var.type)
}
Try the docinfeR package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.