max(joininds)
rep(max(joininds)+(1:5), each=10)
fix(scenariogrid)
floor(3)
floor(3.1)
# Make table of settings:
method <- c("block_forest", "weights_only", "sample_from_blocks",
"one_block_per_split", "block_select_weights", "randomsurvivalforest")
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
cvind <- 1:5
cvfoldind <- 1:5
scenariogrid <- expand.grid(cvind=cvind, dat=dat, cvfoldind=cvfoldind, method=method, stringsAsFactors = FALSE)
scenariogrid <- scenariogrid[,ncol(scenariogrid):1]
set.seed(1234)
seeds <- sample(1000:10000000, size=length(dat)*length(cvind))
scenariogrid$seed <- rep(seeds, times=length(method)*length(cvfoldind))
joininds <- 1:sum(scenariogrid$method=="block_forest" |
scenariogrid$method=="weights_only" |
scenariogrid$method=="sample_from_blocks" |
scenariogrid$method=="one_block_per_split" |
scenariogrid$method=="block_select_weights")
nrep <- floor((nrow(scenariogrid) - max(joininds))/10)
nrep
nrep <- floor((nrow(scenariogrid) - max(joininds))/9)
nrep
(nrow(scenariogrid) - max(joininds))/10
(nrow(scenariogrid) - max(joininds))/9
nrep <- floor((nrow(scenariogrid) - max(joininds))/9)
rep(1:nrep, each=9)
(nrow(scenariogrid) - max(joininds)) - nrep*9
rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*9)
c(rep(1:nrep, each=9), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*9))
ha <- c(rep(1:nrep, each=9), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*9))
length(ha)
joininds
nrow(scenariogrid)
nrep <- floor((nrow(scenariogrid) - max(joininds))/10)
(nrow(scenariogrid) - max(joininds)) - nrep*10
rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)
c(rep(1:nrep, each=9), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10))
c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10))
length(c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)))
sample(c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)))
c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10))
max(joininds) + c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10))
sample(max(joininds) + c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)))
joininds2 <- sample(max(joininds) + c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)))
joininds <- c(joininds, joininds2)
# Make table of settings:
method <- c("block_forest", "weights_only", "sample_from_blocks",
"one_block_per_split", "block_select_weights", "randomsurvivalforest")
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
cvind <- 1:5
cvfoldind <- 1:5
scenariogrid <- expand.grid(cvind=cvind, dat=dat, cvfoldind=cvfoldind, method=method, stringsAsFactors = FALSE)
scenariogrid <- scenariogrid[,ncol(scenariogrid):1]
set.seed(1234)
seeds <- sample(1000:10000000, size=length(dat)*length(cvind))
scenariogrid$seed <- rep(seeds, times=length(method)*length(cvfoldind))
joininds <- 1:sum(scenariogrid$method=="block_forest" |
scenariogrid$method=="weights_only" |
scenariogrid$method=="sample_from_blocks" |
scenariogrid$method=="one_block_per_split" |
scenariogrid$method=="block_select_weights")
nrep <- floor((nrow(scenariogrid) - max(joininds))/10)
joininds2 <- sample(max(joininds) + c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)))
joininds <- c(joininds, joininds2)
scenariogrid$joininds <- joininds
# Make table of settings:
method <- c("block_forest", "weights_only", "sample_from_blocks",
"one_block_per_split", "block_select_weights", "randomsurvivalforest")
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
cvind <- 1:5
cvfoldind <- 1:5
scenariogrid <- expand.grid(cvind=cvind, dat=dat, cvfoldind=cvfoldind, method=method, stringsAsFactors = FALSE)
scenariogrid <- scenariogrid[,ncol(scenariogrid):1]
set.seed(1234)
seeds <- sample(1000:10000000, size=length(dat)*length(cvind))
scenariogrid$seed <- rep(seeds, times=length(method)*length(cvfoldind))
joininds <- 1:sum(scenariogrid$method=="block_forest" |
scenariogrid$method=="weights_only" |
scenariogrid$method=="sample_from_blocks" |
scenariogrid$method=="one_block_per_split" |
scenariogrid$method=="block_select_weights")
nrep <- floor((nrow(scenariogrid) - max(joininds))/10)
joininds2 <- sample(max(joininds) + c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)))
joininds <- c(joininds, joininds2)
scenariogrid$joininds <- joininds
# Randomly permute rows of the table containing the settings.
# This is performed to ensure a comparable computational burdens for
# the jobs to be performed in parallel:
set.seed(1234)
reorderind <- sample(1:nrow(scenariogrid))
scenariogrid <- scenariogrid[reorderind,]
#scenariogrid <- scenariogrid[scenariogrid$joininds %in% c(1,26),]
#scenariogrid$joininds <- 1:2
scenariogrid$settingind <- 1:nrow(scenariogrid)
fix(scenariogrid)
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
patht <- "Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Data"
patht <- "Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Data/"
dati <- paste(patht, dat, sep="")
dati
load(dati[1])
ls()
rm(list=ls());gc()
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
patht <- "Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Data/"
dati <- paste(patht, dat, sep="")
dati
load(dati[1])
ls()
objs <- setdiff(dati, c("dat"))
objs
objs <- setdiff(ls(), c("dat", "dati"))
objs
objs <- setdiff(ls(), c("dat", "dati", "patht", "targetvar"))
objs
dim(cnv)
dim(mirna)
dim(mutation)
dim(rna)
dim(clin)
dim(cnv)
cnv <- cnv[,sort(sample(1:ncol(cnv), size=10))]
dim(mirna)
mirna <- mirna[,sort(sample(1:ncol(mirna), size=10))]
dim(mutation)
mutation <- mutation[,sort(sample(1:ncol(mutation), size=10))]
dim(rna)
rna <- rna[,sort(sample(1:ncol(rna), size=10))]
dim(clin)
# rna <- rna[,sort(sample(1:ncol(rna), size=10))]
gsub(".Rda", "", dati[1])
paste(gsub(".Rda", "", dati[1]), "_small.Rda", collapse="")
paste(gsub(".Rda", "", dati[1]), "_small.Rda", sep="")
save(clin, targetvar, mirna, mutation, cnv, rna, file=paste(gsub(".Rda", "", dati[1]), "_small.Rda", sep=""))
rm(list=ls());gc()
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
patht <- "Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Data/"
dati <- paste(patht, dat, sep="")
dati
i <- 2
load(dati[i])
# ls()
objs <- setdiff(ls(), c("dat", "dati", "patht", "targetvar"))
objs
setdiff(c("clin", "cnv", "i", "mirna", "mutation", "rna") , objs)
i
dim(cnv)
cnv <- cnv[,sort(sample(1:ncol(cnv), size=10))]
dim(mirna)
mirna <- mirna[,sort(sample(1:ncol(mirna), size=10))]
dim(mutation)
mutation <- mutation[,sort(sample(1:ncol(mutation), size=10))]
dim(rna)
rna <- rna[,sort(sample(1:ncol(rna), size=10))]
dim(clin)
# rna <- rna[,sort(sample(1:ncol(rna), size=10))]
save(clin, targetvar, mirna, mutation, cnv, rna, file=paste(gsub(".Rda", "", dati[i]), "_small.Rda", sep=""))
rm(list=ls());gc()
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
patht <- "Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Data/"
dati <- paste(patht, dat, sep="")
dati
i <- 3
load(dati[i])
objs <- setdiff(ls(), c("dat", "dati", "patht", "targetvar"))
objs
setdiff(c("clin", "cnv", "i", "mirna", "mutation", "rna") , objs)
dim(cnv)
cnv <- cnv[,sort(sample(1:ncol(cnv), size=10))]
dim(mirna)
mirna <- mirna[,sort(sample(1:ncol(mirna), size=10))]
dim(mutation)
mutation <- mutation[,sort(sample(1:ncol(mutation), size=10))]
dim(rna)
rna <- rna[,sort(sample(1:ncol(rna), size=10))]
dim(clin)
# rna <- rna[,sort(sample(1:ncol(rna), size=10))]
save(clin, targetvar, mirna, mutation, cnv, rna, file=paste(gsub(".Rda", "", dati[i]), "_small.Rda", sep=""))
rm(list=ls());gc()
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
patht <- "Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Data/"
dati <- paste(patht, dat, sep="")
dati
i <- 4
load(dati[i])
objs <- setdiff(ls(), c("dat", "dati", "patht", "targetvar"))
objs
setdiff(c("clin", "cnv", "i", "mirna", "mutation", "rna") , objs)
dim(cnv)
cnv <- cnv[,sort(sample(1:ncol(cnv), size=10))]
dim(mirna)
mirna <- mirna[,sort(sample(1:ncol(mirna), size=10))]
dim(mutation)
mutation <- mutation[,sort(sample(1:ncol(mutation), size=10))]
dim(rna)
rna <- rna[,sort(sample(1:ncol(rna), size=10))]
dim(clin)
# rna <- rna[,sort(sample(1:ncol(rna), size=10))]
save(clin, targetvar, mirna, mutation, cnv, rna, file=paste(gsub(".Rda", "", dati[i]), "_small.Rda", sep=""))
rm(list=ls());gc()
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
patht <- "Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Data/"
dati <- paste(patht, dat, sep="")
dati
i <- 5
load(dati[i])
objs <- setdiff(ls(), c("dat", "dati", "patht", "targetvar"))
objs
setdiff(c("clin", "cnv", "i", "mirna", "mutation", "rna") , objs)
dim(cnv)
cnv <- cnv[,sort(sample(1:ncol(cnv), size=10))]
dim(mirna)
mirna <- mirna[,sort(sample(1:ncol(mirna), size=10))]
dim(mutation)
mutation <- mutation[,sort(sample(1:ncol(mutation), size=10))]
dim(rna)
rna <- rna[,sort(sample(1:ncol(rna), size=10))]
dim(clin)
# rna <- rna[,sort(sample(1:ncol(rna), size=10))]
save(clin, targetvar, mirna, mutation, cnv, rna, file=paste(gsub(".Rda", "", dati[i]), "_small.Rda", sep=""))
rm(list=ls());gc()
dat <- c("COAD.Rda", "LGG.Rda", "OV.Rda", "PAAD.Rda", "SKCM.Rda", "UCEC.Rda")
patht <- "Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Data/"
dati <- paste(patht, dat, sep="")
dati
i <- 6
load(dati[i])
objs <- setdiff(ls(), c("dat", "dati", "patht", "targetvar"))
objs
setdiff(c("clin", "cnv", "i", "mirna", "mutation", "rna") , objs)
dim(cnv)
cnv <- cnv[,sort(sample(1:ncol(cnv), size=10))]
dim(mirna)
mirna <- mirna[,sort(sample(1:ncol(mirna), size=10))]
dim(mutation)
mutation <- mutation[,sort(sample(1:ncol(mutation), size=10))]
dim(rna)
rna <- rna[,sort(sample(1:ncol(rna), size=10))]
dim(clin)
# rna <- rna[,sort(sample(1:ncol(rna), size=10))]
save(clin, targetvar, mirna, mutation, cnv, rna, file=paste(gsub(".Rda", "", dati[i]), "_small.Rda", sep=""))
rm(list=ls());gc()
# Make table of settings:
method <- c("block_forest", "weights_only", "sample_from_blocks",
"one_block_per_split", "block_select_weights", "randomsurvivalforest")
dat <- c("COAD_small.Rda", "LGG_small.Rda", "OV_small.Rda", "PAAD_small.Rda", "SKCM_small.Rda", "UCEC_small.Rda")
cvind <- 1:5
cvfoldind <- 1:5
scenariogrid <- expand.grid(cvind=cvind, dat=dat, cvfoldind=cvfoldind, method=method, stringsAsFactors = FALSE)
scenariogrid <- scenariogrid[,ncol(scenariogrid):1]
set.seed(1234)
seeds <- sample(1000:10000000, size=length(dat)*length(cvind))
scenariogrid$seed <- rep(seeds, times=length(method)*length(cvfoldind))
joininds <- 1:sum(scenariogrid$method=="block_forest" |
scenariogrid$method=="weights_only" |
scenariogrid$method=="sample_from_blocks" |
scenariogrid$method=="one_block_per_split" |
scenariogrid$method=="block_select_weights")
nrep <- floor((nrow(scenariogrid) - max(joininds))/10)
joininds2 <- sample(max(joininds) + c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)))
joininds <- c(joininds, joininds2)
scenariogrid$joininds <- joininds
# Randomly permute rows of the table containing the settings.
# This is performed to ensure a comparable computational burdens for
# the jobs to be performed in parallel:
set.seed(1234)
reorderind <- sample(1:nrow(scenariogrid))
scenariogrid <- scenariogrid[reorderind,]
#scenariogrid <- scenariogrid[scenariogrid$joininds %in% c(1,26),]
#scenariogrid$joininds <- 1:2
scenariogrid$settingind <- 1:nrow(scenariogrid)
head(scenariogrid)
scenariogrid <- scenariogrid[scenariogrid$cvfoldind==2 & scenariogrid$cvind==2,]
dim(sceanriogrid)
dim(scenariogrid)
scenariogrid
6*365
720000/(6*365)
720000/(6*365*8)
# Make table of settings:
method <- c("block_forest", "weights_only", "sample_from_blocks",
"one_block_per_split", "block_select_weights", "randomsurvivalforest")
dat <- c("COAD_small.Rda", "LGG_small.Rda", "OV_small.Rda", "PAAD_small.Rda", "SKCM_small.Rda", "UCEC_small.Rda")
cvind <- 1:5
cvfoldind <- 1:5
scenariogrid <- expand.grid(cvind=cvind, dat=dat, cvfoldind=cvfoldind, method=method, stringsAsFactors = FALSE)
scenariogrid <- scenariogrid[,ncol(scenariogrid):1]
set.seed(1234)
seeds <- sample(1000:10000000, size=length(dat)*length(cvind))
scenariogrid$seed <- rep(seeds, times=length(method)*length(cvfoldind))
joininds <- 1:sum(scenariogrid$method=="block_forest" |
scenariogrid$method=="weights_only" |
scenariogrid$method=="sample_from_blocks" |
scenariogrid$method=="one_block_per_split" |
scenariogrid$method=="block_select_weights")
nrep <- floor((nrow(scenariogrid) - max(joininds))/10)
joininds2 <- sample(max(joininds) + c(rep(1:nrep, each=10), rep(nrep+1, times=(nrow(scenariogrid) - max(joininds)) - nrep*10)))
joininds <- c(joininds, joininds2)
scenariogrid$joininds <- joininds
# Randomly permute rows of the table containing the settings.
# This is performed to ensure a comparable computational burdens for
# the jobs to be performed in parallel:
set.seed(1234)
reorderind <- sample(1:nrow(scenariogrid))
scenariogrid <- scenariogrid[reorderind,]
#scenariogrid <- scenariogrid[scenariogrid$joininds %in% c(1,26),]
#scenariogrid$joininds <- 1:2
scenariogrid$settingind <- 1:nrow(scenariogrid)
dim(scenariogrid)
sum(length(unique(scenariogrid$joininds)))
sum(length(unique(scenariogrid$joininds)))/25
(1/0.75)*25
765/33
765/131
(765-131)/131
4.83*(2 + 13/60)
(2 + 13/60)*765/131
(15/6)*12.94
24+20
44/32
factorial(5)
0.5^5
0.5^4
0.5^2
0.5^3
qnorm(3)
qnorm(0.8)
qnorm(0.9)
qnorm(0.975)
pm <- c(3, 5, 10)
cvalues <- sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1))
cvalues
pm
cvalues <- sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1))
cvalues
cvalues <- sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1))
cvalues
cvalues <- sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1))
cvalues
cvalues <- sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1))
cvalues
cvalues <- sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1))
cvalues
cvalues <- sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1))
cvalues
pm
cvalues <- sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1))
pm
cvalues
aha <- replicate(1000, sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1)))
dim(aha)
range(aha[3,])
hist(aha[3,])
aha <- replicate(100000, sapply(pm, function(x) sample(c(runif(1, 0, sqrt(x)/x), runif(1, sqrt(x)/x, 1)), size=1)))
hist(aha[3,])
hist(aha[3,], 100)
abline(v=sqrt(10)/10, col=2)
mean(aha[3,] < sqrt(10)/10)
5000000/300000
16*500
aha <- list.dirs("S:/cmmgrp/TCGA_Datasets")
aha
datasets2 <- gsub(".Rda", "", datasets)
datasets <- c("BLCA.Rda", "BRCA.Rda", "CESC.Rda", "COAD.Rda",
"ESCA.Rda", "GBM.Rda", "HNSC.Rda", "KIRC.Rda",
"KIRP.Rda", "LGG.Rda", "LIHC.Rda", "LUAD.Rda",
"LUSC.Rda", "OV.Rda", "PAAD.Rda", "PRAD.Rda",
"READ.Rda", "SARC.Rda", "SKCM.Rda", "STAD.Rda",
"UCEC.Rda")
datasets2 <- gsub(".Rda", "", datasets)
datasets2
lapply(datasets2, function(x) grep(x, aha))
?list.dirs
aha <- list.dirs("S:/cmmgrp/TCGA_Datasets", recursive=FALSE)
lapply(datasets2, function(x) grep(x, aha))
sapply(datasets2, function(x) grep(x, aha))
datanames <- aha[sapply(datasets2, function(x) grep(x, aha))]
datanames
aha <- list.dirs("S:/cmmgrp/TCGA_Datasets", full.names=FALSE, recursive=FALSE)
datanames <- aha[sapply(datasets2, function(x) grep(x, aha))]
datanames
sapply(datanames, function(x) {
y <- strsplit(x, split="_")[[1]]
y <- y[-length(y)]
paste(y, collapse=" ")
})
datanames <- sapply(datanames, function(x) {
y <- strsplit(x, split="_")[[1]]
y <- y[-length(y)]
paste(y, collapse=" ")
})
datanames
names(datanames) <- NULL
datanames
# Make table of settings:
method <- c("block_forest", "weights_only", "sample_from_blocks",
"one_block_per_split", "block_select_weights", "randomsurvivalforest")
dat <- c("BLCA.Rda", "CESC.Rda", "ESCA.Rda", "HNSC.Rda", "LUAD.Rda",
"LUSC.Rda", "PRAD.Rda", "READ.Rda", "SARC.Rda")
cvind <- 1:5
cvfoldind <- 1:5
scenariogrid <- expand.grid(cvind=cvind, dat=dat, cvfoldind=cvfoldind, method=method, stringsAsFactors = FALSE)
scenariogrid <- scenariogrid[,ncol(scenariogrid):1]
set.seed(1234)
seeds <- sample(1000:10000000, size=length(dat)*length(cvind))
scenariogrid$seed <- rep(seeds, times=length(method)*length(cvfoldind))
joininds <- 1:nrow(scenariogrid)
scenariogrid$joininds <- joininds
# Randomly permute rows of the table containing the settings.
# This is performed to ensure a comparable computational burdens for
# the jobs to be performed in parallel:
set.seed(1234)
reorderind <- sample(1:nrow(scenariogrid))
scenariogrid <- scenariogrid[reorderind,]
#scenariogrid <- scenariogrid[scenariogrid$joininds %in% c(1,26),]
#scenariogrid$joininds <- 1:2
scenariogrid$settingind <- 1:nrow(scenariogrid)
# scenariogrid <- scenariogrid[scenariogrid$seed==scenariogrid$seed[1],]
# scenariogrid <- scenariogrid[c(1,2,3,4,6,8),]
# scenariogrid$joininds <- 1:nrow(scenariogrid)
dim(scenariogrid)
ui <- "Network Meta-Analysis Correlates with Analysis of Merged Independent Transcriptome Expression Data"
cat(gsub(" ", "_", ui), "\n")
966/1350
977/1350
985/1350
996/1350
1028/1350
files
files <- list.files("Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Results")
head(files)
files <- grep(".Rda", files, value=TRUE)
head(files)
setwd("Z:/Projects/BlockForests/PaperAnalysis/BlockForests/Results")
file.info(files[1])$ctime
as.numeric(file.info(files[1])$ctime)
times <- sapply(files, function(x) file.info(x)$ctime)
times[which.max(as.numeric(times))]
head(times)
times <- lapply(files, function(x) file.info(x)$ctime)
head(times)
times2 <- unlist(times)
head(times2)
times[[which.max(times2)]]
times[[which.min(times2)]]
filesord <- files[order(times2)]
head(filesord)
indsord <- as.numeric(gsub(".Rda", "", gsub("res", "", filesord)))
head(indsord)
# Make table of settings:
method <- c("block_forest", "weights_only", "sample_from_blocks",
"one_block_per_split", "block_select_weights", "randomsurvivalforest")
dat <- c("BLCA.Rda", "CESC.Rda", "ESCA.Rda", "HNSC.Rda", "LUAD.Rda",
"LUSC.Rda", "PRAD.Rda", "READ.Rda", "SARC.Rda")
cvind <- 1:5
cvfoldind <- 1:5
scenariogrid <- expand.grid(cvind=cvind, dat=dat, cvfoldind=cvfoldind, method=method, stringsAsFactors = FALSE)
scenariogrid <- scenariogrid[,ncol(scenariogrid):1]
set.seed(1234)
seeds <- sample(1000:10000000, size=length(dat)*length(cvind))
scenariogrid$seed <- rep(seeds, times=length(method)*length(cvfoldind))
joininds <- 1:nrow(scenariogrid)
scenariogrid$joininds <- joininds
# Randomly permute rows of the table containing the settings.
# This is performed to ensure a comparable computational burdens for
# the jobs to be performed in parallel:
set.seed(1234)
reorderind <- sample(1:nrow(scenariogrid))
scenariogrid <- scenariogrid[reorderind,]
#scenariogrid <- scenariogrid[scenariogrid$joininds %in% c(1,26),]
#scenariogrid$joininds <- 1:2
scenariogrid$settingind <- 1:nrow(scenariogrid)
# scenariogrid <- scenariogrid[scenariogrid$seed==scenariogrid$seed[1],]
# scenariogrid <- scenariogrid[c(1,2,3,4,6,8),]
# scenariogrid$joininds <- 1:nrow(scenariogrid)
head(scenariogrid)
indssel <- sapply(indsord, function(x) which(scenariogrid$joininds==x))
scen2 <- scenariogrid[indssel,]
dim(scen2)
head(scen2)
head(indsord)
methodsord <- scen2$method
fix(scen2)
sapply(method, function(x) median(scen2$method==x))
sapply(method, function(x) median(which(scen2$method==x)))
sort(sapply(method, function(x) median(which(scen2$method==x)))
)
sort(sapply(method, function(x) median(which(scen2$method==x))))
plot(sort(sapply(method, function(x) median(which(scen2$method==x)))))
plot(sort(sapply(method, function(x) mean(which(scen2$method==x)))))
sort(sapply(method, function(x) mean(which(scen2$method==x))))
sort(sapply(dat, function(x) mean(which(scen2$dat==x))))
plot(sort(sapply(dat, function(x) mean(which(scen2$dat==x)))))
head(scen2)
sort(sapply(method, function(x) sum(scen2$method==x)/sum(scenariogrid$method==x)))
sort(sapply(dat, function(x) sum(scen2$dat==x)/sum(scenariogrid$dat==x)))
dim(scneioarigrd)
dim(scnenariogrid)
dim(scenariogrid)
1042/1350
fix(scen2)
1048/1350
expand.grid(dat=dat, method=method, stringsAsFactors = FALSE)
apply(expand.grid(dat=dat, method=method, stringsAsFactors = FALSE), 1, function(x) sum(scen2$dat==x[1] & scen2$method==x[2]))
range(apply(expand.grid(dat=dat, method=method, stringsAsFactors = FALSE), 1, function(x) sum(scen2$dat==x[1] & scen2$method==x[2])))
apply(mygrid, 1, function(x) sum(scen2$dat==x[1] & scen2$method==x[2]))
mygrid <- expand.grid(dat=dat, method=method, stringsAsFactors = FALSE)
apply(mygrid, 1, function(x) sum(scen2$dat==x[1] & scen2$method==x[2]))
mygrid[order(apply(mygrid, 1, function(x) sum(scen2$dat==x[1] & scen2$method==x[2]))),]
