summary(model_diff)
summary(model_vor_T)
summary(model_beginn)
summary(model_ende)
boxplot(Alter.TX)
boxplot(data$Alter.TX)
# Wegdifferenz (pre-Reha und Reha-Ende)
#---------------
plot(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehaende - data$X6MWD_PreTx) # almost normally distributed outcome
model_diff_pre <- lm(I(data$X6MWD_Rehaende - data$X6MWD_PreTx) ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff_pre)
summary(model_diff)
summary(model_beginn)
summary(model_diff)
summary(model_diff_pre)
summary(model_vor_T)
exp(-0.06003 )
exp(-0.06003 *10)
plot(data$X6MWD_PreTx, data$Alter)
plot(data$X6MWD_PreTx, data$Alter.TX)
plot(data$X6MWD_PreTx, data$Alter.TX)
plot(data$X6MWD_Rehabeginn, data$Alter.TX)
plot(data$X6MWD_Rehaende, data$Alter.TX)
par(mfrow = c(2,2))
plot(data$X6MWD_PreTx, data$Alter.TX)
plot(data$X6MWD_Rehabeginn, data$Alter.TX)
plot(data$X6MWD_Rehaende, data$Alter.TX)
plot(model_beginn)
plot(model_ende)
plot(model_diff_pre)
plot(model_diff_pre)
diff_pre <- data$X6MWD_Rehaende - data$X6MWD_PreTx
plot(diff_pre, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "pearson")
# multivariable model
boxplot(diff_pre) # almost normally distributed outcome
model_diff_pre <- lm(diff_pre ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff_pre)
plot(model_diff_pre)
data[17,]
boxplot(Alter)
boxplot(data$Alter)
setwd("Z:/tmp/Beratung/Weig/Psoasarea/Reha/")
data <- read.csv2("Anthrop_vor_LuTx_IBE_inkl_Reha.csv", nrows = 106)
# only patients with
data <- data[complete.cases(data[, c("X6MWD_Rehabeginn", "X6MWD_Rehaende")]),]
# Wegdifferenz (Reha-Beginn und Reha-Ende)
#---------------
plot(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn) # almost normally distributed outcome
model_diff <- lm(I(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn) ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
diff <- data$X6MWD_Rehaende - data$X6MWD_Rehabeginn
plot(diff, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(diff) # almost normally distributed outcome
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
data[c(50, 20, 88),]
colnames(data)
rownames(data)
rownames(data) %in% c("50", "20", "88")
data[rownames(data) %in% c("50", "20", "88"),]
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_PreTx, data$TPA)
cor.test(data$X6MWD_PreTx, data$TPA, method = "spearman")
cor.test(data$X6MWD_PreTx, data$TPA, method = "pearson")
# multivariable model (zero-inflated)
hist(data$X6MWD_PreTx)
library(pscl)
model_vor_T <- zeroinfl(X6MWD_PreTx ~ TPA + factor(Sex) + Alter.TX + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others) | TPA + Alter.TX +  as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others),
dist = "negbin", data = data)
summary(model_vor_T)
plot(model_vor_T)
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_Rehabeginn, data$TPA)
cor.test(data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehabeginn) # almost normally distributed outcome
model_beginn <- lm(data$X6MWD_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
model_beginn <- lm(data$X6MWD_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + I(Alter^2) + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
model_beginn <- lm(data$X6MWD_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + I(Alter^2) + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
model_beginn <- lm(data$X6MWD_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + I(Alter.TX^2) + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
model_beginn <- lm(data$X6MWD_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
data$X6MWD_Rehabeginn
boxplot(data$X6MWD_Rehabeginn) # almost normally distributed outcome
summary(model_beginn)
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_Rehabeginn, data$TPA)
cor.test(data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehabeginn) # almost normally distributed outcome
model_beginn <- lm(data$X6MWD_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
cor(data$Alter, data$TPA)
cor(data$BMI, data$TPA)
diff <- data$X6MWD_Rehaende - data$X6MWD_Rehabeginn
plot(diff, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(diff) # almost normally distributed outcome
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + BMI + SAD + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
diff <- data$X6MWD_Rehaende - data$X6MWD_Rehabeginn
plot(diff, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(diff) # almost normally distributed outcome
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
data[rownames(data) %in% c("50", "20", "88"),]
cor(TPA, BMI)
cor(data$TPA, data$BMI)
cor.test(data$TPA, data$BMI)
diff <- data$X6MWD_Rehaende - data$X6MWD_Rehabeginn
plot(diff, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(diff) # almost normally distributed outcome
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
data[rownames(data) %in% c("50", "20", "88"),]
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_PreTx, data$TPA)
cor.test(data$X6MWD_PreTx, data$TPA, method = "spearman")
cor.test(data$X6MWD_PreTx, data$TPA, method = "pearson")
# multivariable model (zero-inflated)
hist(data$X6MWD_PreTx)
library(pscl)
model_vor_T <- zeroinfl(X6MWD_PreTx ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others) | TPA + Alter.TX +  as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others),
dist = "negbin", data = data)
summary(model_vor_T)
# multivariable model (zero-inflated)
hist(data$X6MWD_PreTx)
library(pscl)
model_vor_T <- zeroinfl(X6MWD_PreTx ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others) | TPA + BMI + Alter.TX +  as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others),
dist = "negbin", data = data)
summary(model_vor_T)
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_Rehabeginn, data$TPA)
cor.test(data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehabeginn) # almost normally distributed outcome
model_beginn <- lm(data$X6MWD_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_Rehaende, data$TPA)
cor.test(data$X6MWD_Rehaende, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehaende) # almost normally distributed outcome
model_ende <- lm(data$X6MWD_Rehaende ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_ende)
plot(model_ende)
diff_pre <- data$X6MWD_Rehaende - data$X6MWD_PreTx
plot(diff_pre, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "pearson")
# multivariable model
boxplot(diff_pre) # almost normally distributed outcome
model_diff_pre <- lm(diff_pre ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff_pre)
plot(model_diff_pre)
data <- read.csv2("Anthrop_vor_LuTx_IBE_inkl_Reha.csv", nrows = 106)
# only patients with
data <- data[complete.cases(data[, c("X6MWD_Rehabeginn", "X6MWD_Rehaende")]),]
# Wegdifferenz (Reha-Beginn und Reha-Ende)
#---------------
diff <- data$X6MWD_Rehaende - data$X6MWD_Rehabeginn
plot(diff, data$TPA)
# univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(diff) # almost normally distributed outcome
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
data[rownames(data) %in% c("50", "20", "88"),]
# Strecke Reha-Ende
#------------------
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_Rehaende, data$TPA)
cor.test(data$X6MWD_Rehaende, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehaende) # almost normally distributed outcome
model_ende <- lm(data$X6MWD_Rehaende ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_ende)
plot(model_ende)
# Wegdifferenz (pre-Reha und Reha-Ende)
#---------------
diff_pre <- data$X6MWD_Rehaende - data$X6MWD_PreTx
plot(diff_pre, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "pearson")
# multivariable model
boxplot(diff_pre) # almost normally distributed outcome
model_diff_pre <- lm(diff_pre ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff_pre)
plot(model_diff_pre)
# Strecke vor Transplantation
#----------------------------
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_PreTx, data$TPA)
cor.test(data$X6MWD_PreTx, data$TPA, method = "spearman")
cor.test(data$X6MWD_PreTx, data$TPA, method = "pearson")
# multivariable model (zero-inflated)
hist(data$X6MWD_PreTx)
library(pscl)
model_vor_T <- zeroinfl(X6MWD_PreTx ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others) | TPA + BMI + Alter.TX +  as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others),
dist = "negbin", data = data)
summary(model_vor_T)
# Strecke Reha-Beginn
#--------------------
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_Rehabeginn, data$TPA)
cor.test(data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehabeginn) # almost normally distributed outcome
model_beginn <- lm(data$X6MWD_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
# TPA and baseline value (before surgery) associated?
plot(data$X6MWD_Rehaende, data$TPA)
cor.test(data$X6MWD_Rehaende, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende, data$TPA, method = "pearson")
# multivariable model
boxplot(data$X6MWD_Rehaende) # almost normally distributed outcome
model_ende <- lm(data$X6MWD_Rehaende ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_ende)
plot(model_ende)
setwd("Z:/tmp/Beratung/Weig/Psoasarea/Reha/")
data <- read.csv2("Anthrop_vor_LuTx_IBE_inkl_Reha.csv", nrows = 106)
# only patients with
data <- data[complete.cases(data[, c("TLC_Rehabeginn", "TLC_Rehaende")]),]
diff <- data$TLC_Rehaende - data$TLC_Rehabeginn
plot(diff, data$TPA)
# univariate effect
cor.test(diff, data$TPA, method = "spearman")
cor.test(diff, data$TPA, method = "pearson")
# multivariable model
boxplot(diff) # almost normally distributed outcome
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
data[rownames(data) %in% c("50", "20", "88"),]
plot(diff, data$TPA)
boxplot(diff) # almost normally distributed outcome
plot(data$TLC_Rehabeginn, data$TPA)
cor.test(data$TLC_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$TLC_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(data$TLC_Rehabeginn) # almost normally distributed outcome
model_beginn <- lm(data$TLC_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
plot(data$TLC_Rehaende, data$TPA)
cor.test(data$TLC_Rehaende, data$TPA, method = "spearman")
cor.test(data$TLC_Rehaende, data$TPA, method = "pearson")
# multivariable model
boxplot(data$TLC_Rehaende) # almost normally distributed outcome
model_ende <- lm(data$TLC_Rehaende ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_ende)
data$TPA
data <- read.csv2("Anthrop_vor_LuTx_IBE_inkl_Reha.csv", nrows = 106)
# only patients with
data <- data[complete.cases(data[, c("FEV1_Rehabeginn", "FEV1_Rehaende")]),]
data <- read.csv2("Anthrop_vor_LuTx_IBE_inkl_Reha.csv", nrows = 106)
# only patients with
data <- data[complete.cases(data[, c("FEV.1_Rehabeginn", "FEV.1_Rehaende")]),]
data <- read.csv2("Anthrop_vor_LuTx_IBE_inkl_Reha.csv", nrows = 106)
# only patients with
data <- data[complete.cases(data[, c("FEV.1_Rehabeginn", "FEV.1_Rehaende")]),]
diff <- data$FEV.1_Rehaende - data$FEV.1_Rehabeginn
plot(diff, data$TPA)
# univariate effect
cor.test(diff, data$TPA, method = "spearman")
cor.test(diff, data$TPA, method = "pearson")
# multivariable model
boxplot(diff) # almost normally distributed outcome
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
plot(data$FEV.1_Rehabeginn, data$TPA)
cor.test(data$FEV.1_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$FEV.1_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(data$FEV.1_Rehabeginn) # almost normally distributed outcome
model_beginn <- lm(data$FEV.1_Rehabeginn ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_beginn)
plot(model_beginn)
# FEV Reha-Ende
#------------------
# TPA and baseline value (before surgery) associated?
plot(data$FEV.1_Rehaende, data$TPA)
cor.test(data$FEV.1_Rehaende, data$TPA, method = "spearman")
cor.test(data$FEV.1_Rehaende, data$TPA, method = "pearson")
# multivariable model
boxplot(data$FEV.1_Rehaende) # almost normally distributed outcome
model_ende <- lm(data$FEV.1_Rehaende ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_ende)
plot(model_ende)
data$FEV.1_Rehaende
data <- read.csv2("Anthrop_vor_LuTx_IBE_inkl_Reha.csv", nrows = 106)
# only patients with
data <- data[complete.cases(data[, c("X6MWD_Rehabeginn", "X6MWD_Rehaende")]),]
# Wegdifferenz (Reha-Beginn und Reha-Ende)
#---------------
diff <- data$X6MWD_Rehaende - data$X6MWD_Rehabeginn
plot(diff, data$TPA)
# univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_Rehabeginn, data$TPA, method = "pearson")
# multivariable model
boxplot(diff) # almost normally distributed outcome
model_diff <- lm(diff ~ TPA + factor(Sex) + Alter.TX + as.factor(Organ) + BMI + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff)
plot(model_diff)
data[rownames(data) %in% c("50", "20", "88"),]
# Wegdifferenz (pre-Reha und Reha-Ende)
#---------------
diff_pre <- data$X6MWD_Rehaende - data$X6MWD_PreTx
plot(diff_pre, data$TPA)
# no univariate effect
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "spearman")
cor.test(data$X6MWD_Rehaende - data$X6MWD_PreTx, data$TPA, method = "pearson")
# multivariable model
boxplot(diff_pre) # almost normally distributed outcome
model_diff_pre <- lm(diff_pre ~ TPA + factor(Sex) + Alter.TX + BMI + as.factor(Organ) + as.factor(Grunderkrankung.1.ILD.2.COPD.3.CF.4.ReTx.5.Others), data = data)
summary(model_diff_pre)
plot(model_diff_pre)
path <- "cmmgrp/Silke/RFordinal_SJ_GT_ALB/"
path <- "Z:/cmmgrp/Silke/RFordinal_SJ_GT_ALB/"
setwd(paste(path, "reproducible_files/Real_Data_Applications/R_Objects/", sep = "")) # change working directory
load("../R_Code/TRASH/vlbw.sav")
vlbw <- vlbw[,names(vlbw) %in% c("apg1", "twn", "inout", "sex", "race", "delivery", "meth", "lol", "magsulf", "bwt", "toc")]
# convert variables to the correct data format
vlbw$twn     <- as.factor(vlbw$twn)
vlbw$magsulf <- as.factor(vlbw$magsulf)
vlbw$meth    <- as.factor(vlbw$meth )
vlbw$toc     <- as.factor(vlbw$toc)
vlbw <- vlbw[complete.cases(vlbw),]
table(vlbw$apg1)
vlbw <- vlbw[vlbw$apg1 != 0,] # only 2 observations with apgar = 0
table(vlbw$race)
# exclude observations with too sparse levels
vlbw <- vlbw[vlbw$apg1 != 0,]                 # only 2 observations with apgar = 0
vlbw <- vlbw[vlbw$race != "oriental",]        # only 1 observation with race = oriental
vlbw <- vlbw[vlbw$race != "native American",] # only 4 observations with race = native American
vlbw$race <- factor(vlbw$race)
summary(vlbw)
path <- "cmmgrp/Silke/RFordinal_SJ_GT_ALB/"
path <- "Z:/cmmgrp/Silke/RFordinal_SJ_GT_ALB/"
setwd(paste(path, "reproducible_files/Real_Data_Applications/R_Objects/", sep = "")) # change working directory
require(snowfall)
sfInit(parallel =F)
sfLibrary(party)
sfLibrary(ROCR)
# source the code for our novel variable importance measures
sfSource("../../novel_VIMs.R")
# source the functions implementing the simulation studies
sfSource("../R_Code/functions.R")
# obtain very low birth weight data
load("../R_Code/TRASH/vlbw.sav")
# include only prenatal variables
vlbw <- vlbw[,names(vlbw) %in% c("apg1", "twn", "inout", "sex", "race", "delivery", "meth", "lol", "magsulf", "bwt", "toc")]
# convert variables to the correct data format
vlbw$twn     <- as.factor(vlbw$twn)
vlbw$magsulf <- as.factor(vlbw$magsulf)
vlbw$meth    <- as.factor(vlbw$meth)
vlbw$toc     <- as.factor(vlbw$toc)
# exclude observations with missing information
vlbw <- vlbw[complete.cases(vlbw),]
# exclude observations with too sparse levels
vlbw <- vlbw[vlbw$apg1 != 0,]                 # only 2 observations with apgar = 0
vlbw <- vlbw[vlbw$race != "oriental",]        # only 1 observation with race = oriental
vlbw <- vlbw[vlbw$race != "native American",] # only 4 observations with race = native American
vlbw$race <- factor(vlbw$race)
# specify response as nominal variable
vlbw$apg1 <- factor(vlbw$apg1, ordered = FALSE)
pred_perf_vlbw <- sfSapply(1:1, function(z) prediction_performance(seed = z, data = vlbw, yname = "apg1"))
?rowMeans
result <- sapply(1:ncv, function(cv){
tryCatch({ # if there are technical problems in fitting the random forest, return NA
# Fit random forests
# ------------------
res <- rowMeans(sapply(1:k, function(z) tryCatch({perform(z)}, error = function(e) {
return(c(ER.ord = NA, ER.cat = NA,
RPS.ord = NA, RPS.cat = NA))})), na.rm = TRUE)
RFs <- fit_RF(seed = seed,
traindata = data[cvid != cv,],
yname = yname,
scores = scores,
mtry = mtry)
# Compute prediction performance
# ------------------------------
PA_ordinal        <- get_PA(RF = RFs$RF_ordinal, testdata = data[cvid == cv,], yname = yname)
PA_classification <- get_PA(RF = RFs$RF_classification, testdata = data[cvid == cv,], yname = yname)
result <- c(PA_ordinal, PA_classification)
}, error = function(e){return(c(NA, NA, NA, NA))})
names(result) <- c("RPS_RF_ordinal", "ER_RF_ordinal", "RPS_RF_classification", "ER_RF_classification")
return(result)
})
ncv <- 10
result <- sapply(1:ncv, function(cv){
tryCatch({ # if there are technical problems in fitting the random forest, return NA
# Fit random forests
# ------------------
res <- rowMeans(sapply(1:k, function(z) tryCatch({perform(z)}, error = function(e) {
return(c(ER.ord = NA, ER.cat = NA,
RPS.ord = NA, RPS.cat = NA))})), na.rm = TRUE)
RFs <- fit_RF(seed = seed,
traindata = data[cvid != cv,],
yname = yname,
scores = scores,
mtry = mtry)
# Compute prediction performance
# ------------------------------
PA_ordinal        <- get_PA(RF = RFs$RF_ordinal, testdata = data[cvid == cv,], yname = yname)
PA_classification <- get_PA(RF = RFs$RF_classification, testdata = data[cvid == cv,], yname = yname)
result <- c(PA_ordinal, PA_classification)
}, error = function(e){return(c(NA, NA, NA, NA))})
names(result) <- c("RPS_RF_ordinal", "ER_RF_ordinal", "RPS_RF_classification", "ER_RF_classification")
return(result)
})
result <- sapply(1:ncv, function(cv){
tryCatch({ # if there are technical problems in fitting the random forest, return NA
# Fit random forests
# ------------------
res <- rowMeans(sapply(1:k, function(z) tryCatch({perform(z)}, error = function(e) {
return(c(ER.ord = NA, ER.cat = NA,
RPS.ord = NA, RPS.cat = NA))})), na.rm = TRUE)
RFs <- fit_RF(seed = seed,
traindata = data[cvid != cv,],
yname = yname,
scores = scores,
mtry = mtry)
# Compute prediction performance
# ------------------------------
PA_ordinal        <- get_PA(RF = RFs$RF_ordinal, testdata = data[cvid == cv,], yname = yname)
PA_classification <- get_PA(RF = RFs$RF_classification, testdata = data[cvid == cv,], yname = yname)
result <- c(PA_ordinal, PA_classification)
}, error = function(e){result <- c(NA, NA, NA, NA)})
names(result) <- c("RPS_RF_ordinal", "ER_RF_ordinal", "RPS_RF_classification", "ER_RF_classification")
return(result)
})
c(a = NA, b = NA)
result <- sapply(1:ncv, function(cv){
tryCatch({ # if there are technical problems in fitting the random forest, return NA
# Fit random forests
# ------------------
res <- rowMeans(sapply(1:k, function(z) tryCatch({perform(z)}, error = function(e) {
return(c(ER.ord = NA, ER.cat = NA,
RPS.ord = NA, RPS.cat = NA))})), na.rm = TRUE)
RFs <- fit_RF(seed = seed,
traindata = data[cvid != cv,],
yname = yname,
scores = scores,
mtry = mtry)
# Compute prediction performance
# ------------------------------
PA_ordinal        <- get_PA(RF = RFs$RF_ordinal, testdata = data[cvid == cv,], yname = yname)
PA_classification <- get_PA(RF = RFs$RF_classification, testdata = data[cvid == cv,], yname = yname)
result <- c(PA_ordinal, PA_classification)
names(result) <- c("RPS_RF_ordinal", "ER_RF_ordinal", "RPS_RF_classification", "ER_RF_classification")
return(result)
}, error = function(e){return(c(RPS_RF_ordinal = NA, ER_RF_ordinal = NA, RPS_RF_classification = NA, ER_RF_classification = NA))})
})
