5250 random survival forest
2024-05-22
options(repos = c(CRAN = "https://cloud.r-project.org"))
install.packages("Hmisc")##
## The downloaded binary packages are in
## /var/folders/yd/g4p50_293gs40n1_0xp8mz780000gn/T//Rtmpbx8twO/downloaded_packagesdata_train <- Desktop/5250/data/SSC-FIRSTDAY-TRAIN.csv
# install.packages("randomForestSRC")
install.packages("Hmisc")##
## The downloaded binary packages are in
## /var/folders/yd/g4p50_293gs40n1_0xp8mz780000gn/T//Rtmpbx8twO/downloaded_packageslibrary(randomForestSRC)##
## randomForestSRC 3.2.3
##
## Type rfsrc.news() to see new features, changes, and bug fixes.
## library(survival)
library(Hmisc)##
## Attaching package: 'Hmisc'## The following object is masked from 'package:randomForestSRC':
##
## impute## The following objects are masked from 'package:base':
##
## format.pval, unitsdata_train <- read.csv("/Users/youxueren/Desktop/5250/data/SSC-FIRSTDAY-TRAIN.csv", header = TRUE, sep = ",")
data_test <- read.csv("/Users/youxueren/Desktop/5250/data/SSC-FIRSTDAY.csv", header = TRUE, sep = ",")
# correct the data type
data_test$hospital_expire_flag <- as.integer(gsub("\\[|\\]", "", data_test$hospital_expire_flag))
data_train$hospital_expire_flag <- as.integer(gsub("\\[|\\]", "", data_train$hospital_expire_flag))
# clean the dataset
features_to_remove <- c("stay_id","subject_id","hadm_id","LODS_neurologic",
"LODS_cardiovascular","LODS_renal","LODS_pulmonary",
"LODS_hematologic","LODS_hepatic","OASIS_gcs","OASIS_heart_rate",
"OASIS_mean_bp","OASIS_resp_rate","OASIS_mean_bp_1","OASIS_temp","OASIS_urine_output","SOFA")
# Remove features from the data_train
data_train <- data_train[, !(colnames(data_train) %in% features_to_remove)]
data_test <- data_test[, !(colnames(data_test) %in% features_to_remove)]
# Convert the 'admission_type' column to a factor
data_train$admission_type <- factor(data_train$admission_type)
data_test$admission_type <- factor(data_test$admission_type)
length(data_train)## [1] 172# Calculate the number of missing values per column
missing_counts <- sapply(data_train, function(x) sum(is.na(x)))
# Identify columns with more than 80 missing values
columns_to_drop <- names(missing_counts[missing_counts > 50])
# Drop these columns from data_train
data_train <- data_train[, !(colnames(data_train) %in% columns_to_drop)]
# Optionally, print the names of the dropped columns to check which were removed
print(paste("Dropped columns:", paste(columns_to_drop, collapse = ", ")))## [1] "Dropped columns: lactate_min, lactate_max, ph_min, ph_max, so2_min, so2_max, po2_min, po2_max, pco2_min, pco2_max, aado2_min, aado2_max, aado2_calc_min, aado2_calc_max, pao2fio2ratio_min, pao2fio2ratio_max, baseexcess_min, baseexcess_max, bicarbonate_min, bicarbonate_max, totalco2_min, totalco2_max, hematocrit_min, hematocrit_max, hemoglobin_min, hemoglobin_max, carboxyhemoglobin_min, carboxyhemoglobin_max, methemoglobin_min, methemoglobin_max, temperature_min_1, temperature_max_1, chloride_min, chloride_max, calcium_min, calcium_max, glucose_min_1, glucose_max_1, potassium_min, potassium_max, sodium_min, sodium_max, albumin_min, albumin_max, globulin_min, globulin_max, total_protein_min, total_protein_max, abs_basophils_min, abs_basophils_max, abs_eosinophils_min, abs_eosinophils_max, abs_lymphocytes_min, abs_lymphocytes_max, abs_monocytes_min, abs_monocytes_max, abs_neutrophils_min, abs_neutrophils_max, atyps_min, atyps_max, bands_min, bands_max, imm_granulocytes_min, imm_granulocytes_max, metas_min, metas_max, nrbc_min, nrbc_max, d_dimer_min, d_dimer_max, fibrinogen_min, fibrinogen_max, thrombin_min, thrombin_max, alt_min, alt_max, alp_min, alp_max, ast_min, ast_max, amylase_min, amylase_max, bilirubin_total_min, bilirubin_total_max, bilirubin_direct_min, bilirubin_direct_max, bilirubin_indirect_min, bilirubin_indirect_max, ck_cpk_min, ck_cpk_max, ck_mb_min, ck_mb_max, ggt_min, ggt_max, ld_ldh_min, ld_ldh_max, respiration, liver"length(data_train)## [1] 74# Summarize the number of missing values per column
colSums(is.na(data_train))## ICU_length_of_stay hospital_expire_flag
## 0 0
## age admission_type
## 0 0
## number_of_ICD_codes charlson_comorbidity_index
## 0 0
## OASIS_pre_icu_los OASIS_mechvent
## 0 0
## OASIS_elective_surgery heart_rate_min
## 0 3
## heart_rate_max heart_rate_mean
## 3 3
## sbp_min sbp_max
## 3 3
## sbp_mean dbp_min
## 3 3
## dbp_max dbp_mean
## 3 3
## mbp_min mbp_max
## 3 3
## mbp_mean resp_rate_min
## 3 3
## resp_rate_max resp_rate_mean
## 3 3
## temperature_min temperature_max
## 6 6
## temperature_mean spo2_min
## 6 3
## spo2_max spo2_mean
## 3 3
## glucose_min glucose_max
## 7 7
## glucose_mean gcs_min
## 7 6
## gcs_motor gcs_verbal
## 7 6
## gcs_eyes gcs_unable
## 7 6
## hematocrit_min_1 hematocrit_max_1
## 2 2
## hemoglobin_min_1 hemoglobin_max_1
## 2 2
## platelets_min platelets_max
## 2 2
## wbc_min wbc_max
## 2 2
## aniongap_min aniongap_max
## 6 6
## bicarbonate_min_1 bicarbonate_max_1
## 6 6
## bun_min bun_max
## 3 3
## calcium_min_1 calcium_max_1
## 13 13
## chloride_min_1 chloride_max_1
## 6 6
## creatinine_min creatinine_max
## 5 5
## glucose_min_2 glucose_max_2
## 6 6
## sodium_min_1 sodium_max_1
## 6 6
## potassium_min_1 potassium_max_1
## 5 5
## inr_min inr_max
## 25 25
## pt_min pt_max
## 25 25
## ptt_min ptt_max
## 25 25
## coagulation cardiovascular
## 2 3
## cns renal
## 6 0data_train <- na.omit(data_train)
# 查看清理后的数据
print(dim(data_train))## [1] 156 74library(randomForestSRC)
# If you wish to explicitly omit rows with missing data, you can specify:
# the nodesize：the minimum number of cases in a terminal node of the tree. 1/5/10
# ntree： 500，1000，2000
# splitrule = logrank / random
model1 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 1000,splitrule = "logrank", nodesize = 1, importance = TRUE)
model2 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 500,splitrule = "random", nodesize = 1, importance = TRUE)
model3 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 1000,splitrule = "logrank", nodesize = 5, importance = TRUE)
model4 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 500,splitrule = "random", nodesize = 5, importance = TRUE)
model5 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 1000,splitrule = "logrank", nodesize = 10, importance = TRUE)
model6 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 500,splitrule = "random", nodesize = 10, importance = TRUE)
model7 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 2000,splitrule = "logrank", nodesize = 1, importance = TRUE)
model8 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 2000,splitrule = "random", nodesize = 1, importance = TRUE)
model9 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 2000,splitrule = "logrank", nodesize = 5, importance = TRUE)
model10 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 2000,splitrule = "random", nodesize = 5, importance = TRUE)
model11 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 2000,splitrule = "logrank", nodesize = 10, importance = TRUE)
model12 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 2000,splitrule = "random", nodesize = 10, importance = TRUE)
model13 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 500, splitrule = "logrank", nodesize = 1, importance = TRUE)
model14 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 500, splitrule = "logrank", nodesize = 5, importance = TRUE)
model15 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 500, splitrule = "logrank", nodesize = 10, importance = TRUE)
model16 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 1000, splitrule = "random", nodesize = 1, importance = TRUE)
model17 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 1000, splitrule = "random", nodesize = 5, importance = TRUE)
model18 <- rfsrc(Surv(ICU_length_of_stay, hospital_expire_flag) ~ ., data = data_train,
ntree = 1000, splitrule = "random", nodesize = 10, importance = TRUE)
# Print the model summary
print(model1)## Sample size: 156
## Number of deaths: 28
## Number of trees: 1000
## Forest terminal node size: 1
## Average no. of terminal nodes: 37.828
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.18267935
## (OOB) Requested performance error: 0.14990421print(model2)## Sample size: 156
## Number of deaths: 28
## Number of trees: 500
## Forest terminal node size: 1
## Average no. of terminal nodes: 44.178
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.19131456
## (OOB) Requested performance error: 0.12356322print(model3)## Sample size: 156
## Number of deaths: 28
## Number of trees: 1000
## Forest terminal node size: 5
## Average no. of terminal nodes: 13.991
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.15662612
## (OOB) Requested performance error: 0.15277778print(model4)## Sample size: 156
## Number of deaths: 28
## Number of trees: 500
## Forest terminal node size: 5
## Average no. of terminal nodes: 17.692
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.17544176
## (OOB) Requested performance error: 0.15948276print(model5)## Sample size: 156
## Number of deaths: 28
## Number of trees: 1000
## Forest terminal node size: 10
## Average no. of terminal nodes: 9.942
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.15720142
## (OOB) Requested performance error: 0.16187739print(model6)## Sample size: 156
## Number of deaths: 28
## Number of trees: 500
## Forest terminal node size: 10
## Average no. of terminal nodes: 10.044
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.17745728
## (OOB) Requested performance error: 0.15469349print(model7)## Sample size: 156
## Number of deaths: 28
## Number of trees: 2000
## Forest terminal node size: 1
## Average no. of terminal nodes: 37.4185
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.18085924
## (OOB) Requested performance error: 0.14319923print(model8)## Sample size: 156
## Number of deaths: 28
## Number of trees: 2000
## Forest terminal node size: 1
## Average no. of terminal nodes: 43.6025
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.1948914
## (OOB) Requested performance error: 0.13409962print(model9)## Sample size: 156
## Number of deaths: 28
## Number of trees: 2000
## Forest terminal node size: 5
## Average no. of terminal nodes: 14.0305
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.15584295
## (OOB) Requested performance error: 0.15708812print(model10)## Sample size: 156
## Number of deaths: 28
## Number of trees: 2000
## Forest terminal node size: 5
## Average no. of terminal nodes: 17.5445
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.17758012
## (OOB) Requested performance error: 0.15373563print(model11)## Sample size: 156
## Number of deaths: 28
## Number of trees: 2000
## Forest terminal node size: 10
## Average no. of terminal nodes: 10
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.15740813
## (OOB) Requested performance error: 0.16139847print(model12)## Sample size: 156
## Number of deaths: 28
## Number of trees: 2000
## Forest terminal node size: 10
## Average no. of terminal nodes: 10.1655
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.17827986
## (OOB) Requested performance error: 0.14846743print(model13)## Sample size: 156
## Number of deaths: 28
## Number of trees: 500
## Forest terminal node size: 1
## Average no. of terminal nodes: 37.384
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.18240845
## (OOB) Requested performance error: 0.1532567print(model14)## Sample size: 156
## Number of deaths: 28
## Number of trees: 500
## Forest terminal node size: 5
## Average no. of terminal nodes: 13.96
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.15667571
## (OOB) Requested performance error: 0.15948276print(model15)## Sample size: 156
## Number of deaths: 28
## Number of trees: 500
## Forest terminal node size: 10
## Average no. of terminal nodes: 10.086
## No. of variables tried at each split: 9
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: logrank *random*
## Number of random split points: 10
## (OOB) CRPS: 0.1555463
## (OOB) Requested performance error: 0.16475096print(model16)## Sample size: 156
## Number of deaths: 28
## Number of trees: 1000
## Forest terminal node size: 1
## Average no. of terminal nodes: 43.129
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.19641767
## (OOB) Requested performance error: 0.14750958print(model17)## Sample size: 156
## Number of deaths: 28
## Number of trees: 1000
## Forest terminal node size: 5
## Average no. of terminal nodes: 17.539
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.17778849
## (OOB) Requested performance error: 0.1585249print(model18)## Sample size: 156
## Number of deaths: 28
## Number of trees: 1000
## Forest terminal node size: 10
## Average no. of terminal nodes: 9.995
## No. of variables tried at each split: 1
## Total no. of variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## Splitting rule: random
## (OOB) CRPS: 0.17881111
## (OOB) Requested performance error: 0.16810345length(model4$predicted.oob)## [1] 156length(data_train$ICU_length_of_stay)## [1] 156pred <- predict(model2, data_test)
print(pred)## Sample size of test (predict) data: 80
## Number of grow trees: 500
## Average no. of grow terminal nodes: 44.178
## Total no. of grow variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## CRPS: 0.27449873
## Requested performance error: 0.19720768vi <- vimp(model2, cause = 1) # Adjust 'cause' if you have competing risks
print(vi)## Sample size of test (predict) data: 156
## Number of grow trees: 500
## Average no. of grow terminal nodes: 44.178
## Total no. of grow variables: 72
## Resampling used to grow trees: swor
## Resample size used to grow trees: 99
## Analysis: RSF
## Family: surv
## CRPS: 0.19131456
## Requested performance error: 0.12356322plot(model2)
##
## Importance Relative Imp
## coagulation 0.0331 1.0000
## potassium_max_1 0.0240 0.7247
## gcs_motor 0.0236 0.7135
## inr_min 0.0231 0.6982
## creatinine_max 0.0201 0.6063
## temperature_min 0.0184 0.5553
## calcium_min_1 0.0179 0.5405
## bicarbonate_max_1 0.0161 0.4863
## dbp_min 0.0159 0.4790
## bicarbonate_min_1 0.0158 0.4782
## ptt_max 0.0155 0.4669
## inr_max 0.0153 0.4605
## sodium_max_1 0.0147 0.4430
## OASIS_mechvent 0.0146 0.4405
## aniongap_min 0.0146 0.4397
## glucose_max_2 0.0145 0.4375
## ptt_min 0.0143 0.4320
## gcs_eyes 0.0142 0.4295
## renal 0.0136 0.4097
## charlson_comorbidity_index 0.0134 0.4046
## cardiovascular 0.0132 0.3975
## pt_min 0.0126 0.3812
## pt_max 0.0126 0.3804
## heart_rate_max 0.0125 0.3786
## cns 0.0118 0.3567
## number_of_ICD_codes 0.0117 0.3529
## heart_rate_min 0.0116 0.3495
## mbp_min 0.0114 0.3452
## glucose_max 0.0113 0.3424
## temperature_mean 0.0112 0.3397
## gcs_verbal 0.0108 0.3254
## admission_type 0.0106 0.3208
## resp_rate_mean 0.0101 0.3051
## hemoglobin_min_1 0.0095 0.2867
## temperature_max 0.0089 0.2679
## gcs_unable 0.0089 0.2676
## dbp_max 0.0081 0.2445
## hematocrit_min_1 0.0081 0.2445
## resp_rate_min 0.0078 0.2340
## glucose_min_2 0.0077 0.2331
## sodium_min_1 0.0077 0.2314
## hematocrit_max_1 0.0074 0.2241
## hemoglobin_max_1 0.0069 0.2073
## platelets_max 0.0067 0.2027
## potassium_min_1 0.0066 0.1993
## gcs_min 0.0065 0.1954
## spo2_mean 0.0064 0.1923
## chloride_max_1 0.0063 0.1912
## sbp_mean 0.0063 0.1897
## platelets_min 0.0060 0.1810
## chloride_min_1 0.0056 0.1698
## bun_min 0.0055 0.1669
## spo2_max 0.0055 0.1667
## creatinine_min 0.0054 0.1618
## sbp_min 0.0051 0.1551
## sbp_max 0.0049 0.1473
## resp_rate_max 0.0039 0.1163
## aniongap_max 0.0038 0.1136
## calcium_max_1 0.0031 0.0924
## mbp_max 0.0029 0.0865
## glucose_mean 0.0026 0.0775
## wbc_min 0.0025 0.0760
## glucose_min 0.0024 0.0724
## wbc_max 0.0017 0.0519
## OASIS_elective_surgery 0.0015 0.0463
## heart_rate_mean 0.0015 0.0455
## age 0.0006 0.0175
## dbp_mean 0.0002 0.0047
## OASIS_pre_icu_los -0.0002 -0.0054
## bun_max -0.0009 -0.0283
## spo2_min -0.0019 -0.0570
## mbp_mean -0.0050 -0.1522# Print the C-index from the model output
# Print the C-index from the model output
print(model4$c.index)## NULLget.cindex(time = data_train$ICU_length_of_stay, censoring = data_train$hospital_expire_flag, predicted = model4$predicted.oob)## [1] 0.1594828# less brier score would be better : the 0.25 is the threshold
# inverse probability of censoring weights (IPCW)
## obtain Brier score using KM and RSF censoring distribution estimators
bs.km <- get.brier.survival(model10, cens.mode = "km")$brier.score
bs.rsf <- get.brier.survival(model1, cens.mode = "rfsrc")$brier.score
## plot the brier score
plot(bs.km, type = "s", col = 2)
lines(bs.rsf, type ="s", col = 4)
legend("bottomright", legend = c("cens.model = km", "cens.model = rfsrc"), fill = c(2,4))
## obtain Brier score using KM and RSF censoring distribution estimators
bs.km <- get.brier.survival(model2, cens.mode = "km")$brier.score
bs.rsf <- get.brier.survival(model2, cens.mode = "rfsrc")$brier.score
## plot the brier score
plot(bs.km, type = "s", col = 2)
lines(bs.rsf, type ="s", col = 4)
legend("bottomright", legend = c("cens.model = km", "cens.model = rfsrc"), fill = c(2,4))
oo <- subsample(model2, verbose = FALSE)
# take a delete-d-jackknife procedure for example
vimpCI <- extract.subsample(oo)$var.jk.sel.Z
vimpCI## lower mean upper pvalue
## age -0.711158366 0.05785715 0.8268727 0.441385040
## admission_type -0.224768417 1.06258652 2.3499415 0.052857166
## number_of_ICD_codes -0.018565728 1.16887850 2.3563227 0.026845581
## charlson_comorbidity_index -0.130078532 1.34002666 2.8101319 0.037005809
## OASIS_pre_icu_los -0.808145741 -0.01798453 0.7721767 0.517790882
## OASIS_mechvent 0.204930269 1.45902566 2.7131210 0.011296746
## OASIS_elective_surgery -0.072330368 0.15342073 0.3791718 0.091431082
## heart_rate_min -0.322409468 1.15762704 2.6376636 0.062636983
## heart_rate_max -0.203138805 1.25396966 2.7110781 0.045828511
## heart_rate_mean -0.585535221 0.15072359 0.8869824 0.344123554
## sbp_min -0.611082482 0.51373079 1.6385441 0.185349358
## sbp_max -0.445829818 0.48785888 1.4215476 0.152895227
## sbp_mean -0.409776474 0.62829603 1.6663685 0.117757214
## dbp_min 0.026284747 1.58653414 3.1467835 0.023131680
## dbp_max -0.261119224 0.80994327 1.8810058 0.069152135
## dbp_mean -0.754606047 0.01556332 0.7857327 0.484203504
## mbp_min -0.078057590 1.14348754 2.3650327 0.033273526
## mbp_max -0.634169822 0.28643521 1.2070402 0.270990766
## mbp_mean -1.727919107 -0.50412035 0.7196784 0.790273166
## resp_rate_min -0.194157412 0.77508857 1.7443346 0.058516694
## resp_rate_max -0.404644962 0.38504532 1.1747356 0.169622165
## resp_rate_mean -0.160092910 1.01048107 2.1810550 0.045331512
## temperature_min 0.531514810 1.83933876 3.1471627 0.002921037
## temperature_max -0.235103172 0.88748334 2.0100699 0.060632114
## temperature_mean -0.059310725 1.12498755 2.3092858 0.031315134
## spo2_min -1.140857564 -0.18891508 0.7630274 0.651346824
## spo2_max -0.266690153 0.55216617 1.3710225 0.093145547
## spo2_mean -0.284605865 0.63676575 1.5581374 0.087781522
## glucose_min -0.686294679 0.23987069 1.1660361 0.305860748
## glucose_max -0.198841522 1.13401022 2.4668620 0.047700886
## glucose_mean -0.567342635 0.25685072 1.0810441 0.270665652
## gcs_min -0.121697265 0.64723286 1.4161630 0.049495626
## gcs_motor 0.548969537 2.36330928 4.1776490 0.005340045
## gcs_verbal 0.031789302 1.07777134 2.1237534 0.021716036
## gcs_eyes -0.030248312 1.42256222 2.8753728 0.027482268
## gcs_unable -0.167607663 0.88618464 1.9399769 0.049652947
## hematocrit_min_1 -0.249883844 0.80989425 1.8696723 0.067089186
## hematocrit_max_1 -0.378816330 0.74221222 1.8632408 0.097202816
## hemoglobin_min_1 -0.015117252 0.94969492 1.9145071 0.026849662
## hemoglobin_max_1 -0.234033875 0.68650078 1.6070354 0.071915992
## platelets_min -0.536528336 0.59960964 1.7357476 0.150476212
## platelets_max -0.285068205 0.67149365 1.6280555 0.084431168
## wbc_min -0.444254367 0.25183302 0.9479204 0.239136522
## wbc_max -0.622000487 0.17189129 0.9657831 0.335649428
## aniongap_min 0.021470668 1.45633539 2.8912001 0.023334491
## aniongap_max -0.598614569 0.37626793 1.3511504 0.224683042
## bicarbonate_min_1 0.081499705 1.58371029 3.0859209 0.019400162
## bicarbonate_max_1 -0.006232367 1.61072428 3.2276809 0.025444801
## bun_min -0.174734931 0.55286420 1.2804633 0.068207715
## bun_max -0.970394065 -0.09383656 0.7827210 0.583094571
## calcium_min_1 0.164646464 1.79022285 3.4157992 0.015445309
## calcium_max_1 -0.585419927 0.30603275 1.1974854 0.250521636
## chloride_min_1 -0.350978488 0.56237396 1.4757264 0.113754805
## chloride_max_1 -0.349372442 0.63325732 1.6158871 0.103276277
## creatinine_min -0.330205420 0.53591115 1.4020277 0.112616311
## creatinine_max -0.198661111 2.00819836 4.2150578 0.037250168
## glucose_min_2 -0.276691374 0.77217635 1.8210441 0.074520749
## glucose_max_2 -0.116827107 1.44904426 3.0149156 0.034859323
## sodium_min_1 -0.453688874 0.76637606 1.9864410 0.109135448
## sodium_max_1 0.112407453 1.46742951 2.8224516 0.016895575
## potassium_min_1 -0.339423457 0.66000202 1.6594275 0.097776798
## potassium_max_1 -0.254299358 2.40041510 5.0551296 0.038179364
## inr_min 0.351621372 2.31244785 4.2732743 0.010404539
## inr_max 0.217346082 1.52519756 2.8330490 0.011136441
## pt_min 0.022043915 1.26253253 2.5030211 0.023032718
## pt_max 0.022753943 1.26004100 2.4973281 0.022966521
## ptt_min -0.085483093 1.43081891 2.9471209 0.032195061
## ptt_max 0.056449812 1.54649490 3.0365400 0.020964886
## coagulation 0.891196474 3.31215438 5.7331123 0.003665077
## cardiovascular -0.077783164 1.31671692 2.7112170 0.032110675
## cns 0.033600089 1.18129283 2.3289856 0.021829798
## renal -0.159118190 1.35693739 2.8729930 0.039693393
## signif
## age FALSE
## admission_type FALSE
## number_of_ICD_codes FALSE
## charlson_comorbidity_index FALSE
## OASIS_pre_icu_los FALSE
## OASIS_mechvent TRUE
## OASIS_elective_surgery FALSE
## heart_rate_min FALSE
## heart_rate_max FALSE
## heart_rate_mean FALSE
## sbp_min FALSE
## sbp_max FALSE
## sbp_mean FALSE
## dbp_min TRUE
## dbp_max FALSE
## dbp_mean FALSE
## mbp_min FALSE
## mbp_max FALSE
## mbp_mean FALSE
## resp_rate_min FALSE
## resp_rate_max FALSE
## resp_rate_mean FALSE
## temperature_min TRUE
## temperature_max FALSE
## temperature_mean FALSE
## spo2_min FALSE
## spo2_max FALSE
## spo2_mean FALSE
## glucose_min FALSE
## glucose_max FALSE
## glucose_mean FALSE
## gcs_min FALSE
## gcs_motor TRUE
## gcs_verbal TRUE
## gcs_eyes FALSE
## gcs_unable FALSE
## hematocrit_min_1 FALSE
## hematocrit_max_1 FALSE
## hemoglobin_min_1 FALSE
## hemoglobin_max_1 FALSE
## platelets_min FALSE
## platelets_max FALSE
## wbc_min FALSE
## wbc_max FALSE
## aniongap_min TRUE
## aniongap_max FALSE
## bicarbonate_min_1 TRUE
## bicarbonate_max_1 FALSE
## bun_min FALSE
## bun_max FALSE
## calcium_min_1 TRUE
## calcium_max_1 FALSE
## chloride_min_1 FALSE
## chloride_max_1 FALSE
## creatinine_min FALSE
## creatinine_max FALSE
## glucose_min_2 FALSE
## glucose_max_2 FALSE
## sodium_min_1 FALSE
## sodium_max_1 TRUE
## potassium_min_1 FALSE
## potassium_max_1 FALSE
## inr_min TRUE
## inr_max TRUE
## pt_min TRUE
## pt_max TRUE
## ptt_min FALSE
## ptt_max TRUE
## coagulation TRUE
## cardiovascular FALSE
## cns TRUE
## renal FALSEsignificant_features <- vimpCI[vimpCI$signif == TRUE, ]
# Print the significant features
print(significant_features)## lower mean upper pvalue signif
## OASIS_mechvent 0.20493027 1.459026 2.713121 0.011296746 TRUE
## dbp_min 0.02628475 1.586534 3.146784 0.023131680 TRUE
## temperature_min 0.53151481 1.839339 3.147163 0.002921037 TRUE
## gcs_motor 0.54896954 2.363309 4.177649 0.005340045 TRUE
## gcs_verbal 0.03178930 1.077771 2.123753 0.021716036 TRUE
## aniongap_min 0.02147067 1.456335 2.891200 0.023334491 TRUE
## bicarbonate_min_1 0.08149970 1.583710 3.085921 0.019400162 TRUE
## calcium_min_1 0.16464646 1.790223 3.415799 0.015445309 TRUE
## sodium_max_1 0.11240745 1.467430 2.822452 0.016895575 TRUE
## inr_min 0.35162137 2.312448 4.273274 0.010404539 TRUE
## inr_max 0.21734608 1.525198 2.833049 0.011136441 TRUE
## pt_min 0.02204392 1.262533 2.503021 0.023032718 TRUE
## pt_max 0.02275394 1.260041 2.497328 0.022966521 TRUE
## ptt_max 0.05644981 1.546495 3.036540 0.020964886 TRUE
## coagulation 0.89119647 3.312154 5.733112 0.003665077 TRUE
## cns 0.03360009 1.181293 2.328986 0.021829798 TRUE# Confidence Intervals for VIMP
plot.subsample(oo)
plot.variable(model4, xvar.names = "sbp_min", partial = TRUE)
plot.variable(model4, xvar.names = "temperature_min", partial = TRUE)
plot.variable(model4, xvar.names = "temperature_mean", partial = TRUE)
plot.variable(model4, xvar.names = "gcs_motor", partial = TRUE)
plot.variable(model4, xvar.names = "aniongap_min", partial = TRUE)
plot.variable(model4, xvar.names = "bicarbonate_max_1", partial = TRUE)
plot.variable(model4, xvar.names = "inr_min", partial = TRUE)
plot.variable(model4, xvar.names = "pt_max", partial = TRUE)
plot.variable(model4, xvar.names = "inr_max", partial = TRUE)