#!/usr/bin/Rscript   
getwd()
setwd("~/git/mosaic_2020/")
getwd()
############################################################
# TASK: unpaired (time) analysis of clinical data
# data: clincial data of flu positive adult patients
# group: obesity
############################################################
#my_sample_type = "npa"

#=============
# Input
#=============
source("data_extraction_formatting_clinical.R")

#=============
# Output
#=============
outfile_name_clinical = paste0("flu_stats_clinical_unpaired.csv")
outfile_clinical_unpaired = paste0(outdir_stats, outfile_name_clinical)
outfile_clinical_unpaired 

########################################################################
# Unpaired stats for clinical data b/w groups: wilcoxon UNpaired analysis 
# No correction required
########################################################################

numerical_cols = c("age"
                   #, "vl_pfu_ul_npa1"
                   , "los"
                   , "onset2final"
                   , "onsfindeath"
                   , "onset_2_initial"
                   , "o2_sat_admis")

metadata_cols = c("mosaic", "obesity")

clinical_df_numerical = clinical_df[, c(metadata_cols, numerical_cols)]

pivot_cols = metadata_cols
#pivot_cols = metadata_cols[!meta_data_cols%in%cols_to_omit];pivot_cols 
expected_rows_clinical_lf = nrow(clinical_df_numerical) * (length(clinical_df_numerical) - length(pivot_cols)); expected_rows_clinical_lf


keycol <- "clinical_params"
valuecol <- "value"
gathercols <- c("age", "los", "onset2final", "onsfindeath", "onset_2_initial", "o2_sat_admis")

clinical_lf = gather_(clinical_df_numerical, keycol, valuecol, gathercols)

if( nrow(clinical_lf) == expected_rows_clinical_lf){
  cat("PASS: long format data created successfully"
      , "\nnrow:", nrow(clinical_lf)
      , "\nncol:", ncol(clinical_lf))
}

#====================
# unpaired: clinical
#====================
sum(is.na(clinical_lf$value))

foo = clinical_lf[which(is.na(clinical_lf$value)),]

clinical_lf_comp = clinical_lf[-which(is.na(clinical_lf$value)),]

stats_un_clinical = compare_means(value~obesity
                            , group.by = "clinical_params"
                            , data = clinical_lf 
                            #, data = clinical_lf_comp
                            , paired = FALSE)


stat_df <- clinical_lf %>%
  group_by(clinical_params) %>%
  wilcox_test(value ~ obesity, paired = F) %>%
  add_significance("p")
stat_df$p_format = round(stat_df$p, digits = 3)

#----------------------------------------
# calculate n_obs for each clinical param: Overall
#----------------------------------------
#n_t1 = data.frame(table(lf_t1_comp$mediator))
n_all = data.frame(table(clinical_lf$clinical_params))
colnames(n_all) = c("clinical_params", "n")
n_all$clinical_params = as.character(n_all$clinical_params)

n_gp_lf = data.frame(table(clinical_lf$clinical_params, clinical_lf$obesity))
n_gp = spread(n_gp_lf, "Var2", "Freq"); n_gp
colnames(n_gp)
colnames(n_gp) = c("clinical_params"
                   , paste0("n_gp", colnames(n_gp)[2])
                   , paste0("n_gp", colnames(n_gp)[3]))

n_gp$clinical_params = as.character(n_gp$clinical_params)

n_all_gp  = merge(n_all, n_gp
                  , by = intersect( names(n_all), names(n_gp) )
                  , all = T)

#----------------------------------------
# calculate n_obs for each clinical param: complete cases
#----------------------------------------
n_comp = data.frame(table(clinical_lf_comp$clinical_params))
colnames(n_comp) = c("clinical_params", "n_complete")
n_comp$clinical_params = as.character(n_comp$clinical_params)
n_comp

n_gp_comp_lf = data.frame(table(clinical_lf_comp$clinical_params, clinical_lf_comp$obesity)); n_gp_comp_lf
n_gp_comp = spread(n_gp_comp_lf, "Var2", "Freq"); n_gp_comp
colnames(n_gp_comp)
colnames(n_gp_comp) = c("clinical_params"
                   , paste0("n_complete_gp", colnames(n_gp_comp)[2])
                   , paste0("n_complete_gp", colnames(n_gp_comp)[3]))


n_comp_gp  = merge(n_comp, n_gp_comp
                  , by = intersect( names(n_comp), names(n_gp_comp))
                  , all = T)

merge_cols = intersect(names(n_all_gp), names(n_comp_gp)); merge_cols

n_df = merge(n_all_gp, n_comp_gp, by = merge_cols, all = T); n_df

#==================================
# Merge: merge stats + n_obs df
#===================================
merging_cols = intersect(names(stats_un_clinical), names(n_df)); merging_cols
if (all(n_df$clinical_params%in%stats_un_clinical$clinical_params)) {
  cat("PASS: merging stats and n_obs on column/s:", merging_cols)
  stats_un_clinical = merge(stats_un_clinical, n_df, by = merging_cols, all = T)
  cat("\nsuccessfull merge:"
      , "\nnrow:", nrow(stats_un_clinical)
      , "\nncol:", ncol(stats_un_clinical))
}else{
  nf = n_df$clinical_params[!n_df$clinical_params%in%stats_un_clinical$clinical_params]
  stats_un_clinical = merge(stats_un_clinical, n_df, by = merging_cols, all.y = T)
  cat("\nMerged with caution:"
      , "\nnrows mismatch:", nf
      , "\nnot found in stats possibly due to all obs being missing"
      , "\nintroduced NAs for:", nf
      , "\nnrow:", nrow(stats_un_clinical)
      , "\nncol:", ncol(stats_un_clinical))
}

#######################################################################
#=================
# formatting df
#=================
# delete: unnecessary column
stats_clinical_df = subset(stats_un_clinical, select = -c(.y.,p.adj))

# add: reflect stats method correctly i.e paired or unpaired
# incase there are NA due to LLODs, the gsub won't work!
#stats_clinical_df$method = gsub("Wilcoxon", "Wilcoxon_unpaired", stats_clinical_df$method)
stats_clinical_df$method = "wilcoxon unpaired"
stats_clinical_df$method

# reorder columns
print("preparing to reorder columns...")
colnames(stats_clinical_df)
my_col_order2 = c("clinical_params"
                  , "method"
                  , "group1"
                  , "group2"
                  , "n"
                  , "n_gp0"
                  , "n_gp1"
                  , "n_complete"
                  , "n_complete_gp0"
                  , "n_complete_gp1"
                  , "p"
                  , "p.format"
                  , "p.signif")

if( length(my_col_order2) == ncol(stats_clinical_df) && (all(my_col_order2%in%colnames(stats_clinical_df))) ){
  print("PASS: Reordering columns...")
  stats_clinical_df_f = stats_clinical_df[, my_col_order2]
  print("Successful: column reordering")
  print("formatted df called:'stats_clinical_df_f'")
  cat('\nformatted df has the following dimensions\n')
  print(dim(stats_clinical_df_f ))
} else{
  cat(paste0("FAIL:Cannot reorder columns, length or names mismatch"
             , "\nExpected column order for: ", ncol(stats_clinical_df) # FIXME: can handle better!
             , "\nGot:", length(my_col_order2)
             , "\nElse check colnames to see if they exist in both"))
  quit()
}  
# assign nice column names like replace "." with "_"
colnames(stats_clinical_df_f) = c("clinical_params"
                                   , "method"
                                   , "group1"
                                   , "group2"
                                   , "n"
                                   , "n_gp0"
                                   , "n_gp1"
                                   , "n_complete"
                                   , "n_complete_gp0"
                                   , "n_complete_gp1"
                                   , "p"
                                   , "p_format"
                                   , "p_signif")

colnames(stats_clinical_df_f)
########################################################################
#******************
# write output file
#******************
cat("UNpaired stats for clinical data for groups in:", outfile_clinical_unpaired)
#write.csv(stats_clinical_df_f, outfile_clinical_unpaired, row.names = FALSE)