We sample over distributions of replicates

In [1]:

    

library(data.table)
library(foreach)
library(doParallel)


    

    




Loading required package: iterators
Loading required package: parallel

In [2]:

    

GBD <- read.table("../Data/DALY_YLL_deaths_per_region_and_27_diseases_2005.txt")
GBD <- GBD[order(GBD$metr,GBD$Region,GBD$Disease),]


    

In [3]:

    

#Burden of all diseases in all regions for each metric of burden
GBD <- rbind(GBD,data.frame(metr=levels(GBD$metr),
           Region="All",
           Disease="all",
           burden=tapply(GBD$burden[GBD$Region=="All"],GBD$metr[GBD$Region=="All"],sum)))
rownames(GBD) <- NULL
GBD <- GBD[order(GBD$metr,GBD$Region,GBD$Disease),]

#Burden in Non-HI
DNHI <- GBD[GBD$Region=="All",]
DNHI$burden <- DNHI$burden - GBD$burden[GBD$Region=="High-income"]
DNHI$Region <- "Non-HI"
GBD <- rbind(GBD,DNHI)
GBD$Region <- as.factor(as.character(GBD$Region))
GBD <- GBD[order(GBD$metr,GBD$Region,GBD$Disease),]

#Local proportions of burden across diseases for each metric of burden
GBD$Prp <- 100*GBD$burden/rep(GBD$burden[GBD$Disease=="all"],as.numeric(table(paste(GBD$metr,GBD$Region))))

GBD <- data.table(GBD)


    

In [4]:

    

Mgbd <- read.table("../Data/27_gbd_groups.txt")
sms <- list.files("/media/igna/Elements/HotelDieu/Cochrane/MappingRCTs_vs_Burden/Replicates/Metrics_over_repl/")
dis <- as.numeric(substr(sms,25,nchar(sms)-4))
dis <- dis[dis!=0]


    

In [5]:

    

#We upload all replicates
#We will measure alignment only over diseases for which we have replicates

L <- list()
for(i in 1:length(dis)){
    k <- dis[i]
    DF <- fread(paste(c("/media/igna/Elements/HotelDieu/Cochrane/MappingRCTs_vs_Burden/Replicates/Metrics_over_repl/Metrics_over_replicates_",
                    as.character(k),".txt"),collapse=""))

    regs <- unique(DF$Region)
    L[[i]] <- data.table(
    Disease=Mgbd$x[k],
    Region=DF$Region[DF$Dis=="dis"],
    Prop_RCTs=100*DF$RCTs[DF$Dis=="dis"]/DF$RCTs[DF$Dis=="all"],
    Prop_Patients=100*DF$Patients[DF$Dis=="dis"]/DF$Patients[DF$Dis=="all"],
    sim_nb = rep(1:((nrow(DF)/length(regs))/2),each=length(regs)))
    
    }

DT <- rbindlist(L)

#replicate per disease
DT$k <- paste(DT$Disease,DT$sim_nb)
#nb replicates available per disease
NS <- DT[,max(sim_nb),by="Disease"]


    

In [6]:

    

NK <- 10000
set.seed(1234)
#For each disease, we sample 10,000 replicates, giving local proportions of research
sims <- do.call('rbind',tapply(NS$V1,NS$Disease,function(x){sample(1:x,NK,replace=TRUE)}))


    

In [7]:

    

cl<-makeCluster(4)
registerDoParallel(cl)
t0 <- proc.time()
A <- foreach(k = 1:NK, .packages="data.table") %dopar% {
    x <- sims[,k]
    dtt <- merge(GBD,DT[DT$k%in%paste(rownames(sims),x),])
    dtt[,.(sum(abs(Prp[Prp>=Prop_RCTs]-Prop_RCTs[Prp>=Prop_RCTs])),
       sum(abs(Prp[Prp>=2*Prop_RCTs]/2-Prop_RCTs[Prp>=2*Prop_RCTs])),
       sum(abs(Prp[Prp>=Prop_Patients]-Prop_Patients[Prp>=Prop_Patients])),
       sum(abs(Prp[Prp>=2*Prop_Patients]/2-Prop_RCTs[Prp>=2*Prop_Patients]))),
        by=.(metr,Region)]
    }
stopCluster(cl)
((proc.time() - t0)/60)


    

    






      user     system    elapsed 
 0.1820333  0.0120500 12.7565333 

In [8]:

    

Al <- rbindlist(A)


    

In [9]:

    

head(Al)


    

    






metr
Region
V1
V2
V3
V4

	
1
daly            
All             
30.1461261866829
8.53307379502023
22.7850337488825
3.60760939635803
	
2
death           
All             
41.5116287905914
6.61494087229813
29.1067795691509
1.34425726601647
	
3
yld             
All             
37.9710976667447
7.70993610323922
43.7843448407474
6.71194328228305
	
4
yll             
All             
45.9486562237184
14.5205170931185
34.3092877040155
11.9746418906407
	
5
daly                                            
Central Europe, Eastern Europe, and Central Asia
31.0720417002882                                
6.00432894281371                                
25.1423527596846                                
1.00476603628302                                
	
6
death                                           
Central Europe, Eastern Europe, and Central Asia
53.2621863102998                                
19.7299773036951                                
48.5961498250461                                
19.7299773036951                                

In [10]:

    

AUI <- Al[,lapply(.SD,function(x){quantile(x,probs=c(0.025,0.5,0.975))}),by=.(metr,Region),.SDcols=c("V1","V2","V3","V4")]
AUI$UI <- rep(c("low","med","up"),times=length(unique(paste(AUI$metr,AUI$Region))))
AUI <- AUI[,c("UI","Region","metr","V1","V2","V3","V4")]
names(AUI) <- c("UI","Region","metr","RCTs_fill","RCTs_nogap","Patients_fill","Patients_nogap")


    

In [11]:

    

write.table(AUI,"../Data/Alignment_ratios_within_regions_across_diseases_wt_sims_patients_metrs_burdens.txt")


    

In [ ]: