| Title: | An R Package for Biomarkers Analysis in Precision Medicine |
|---|---|
| Description: | Provides functions for training extreme gradient boosting model using propensity score A-learning and weight-learning methods. For further details, see Liu et al. (2024) <doi:10.1093/bioinformatics/btae592>. |
| Authors: | Zihuan Liu [aut, cre], Yan Sun [aut], Xin Huang [aut] |
| Maintainer: | Zihuan Liu <[email protected]> |
| License: | GPL-3 |
| Version: | 1.0.2 |
| Built: | 2024-11-05 06:44:48 UTC |
| Source: | CRAN |
This function provides a summary of categorical variables in a dataset.
cat_summary( yvar, yname, xvars, xname.list, data, yvar.display = yvar, xvars.display = xvars )cat_summary( yvar, yname, xvars, xname.list, data, yvar.display = yvar, xvars.display = xvars )
yvar |
Name of the variable for summary. |
yname |
A vector of ordered y values. |
xvars |
Names of the variables for grouping. |
xname.list |
A list (same order as xvars) of ordered x values for each xvar. |
data |
The dataset. |
yvar.display |
Display name for yvar. |
xvars.display |
Display name for xvars. |
A list containing the contingency table, frequency table, and percentage table.
# Load a sample dataset data <- data.frame( outcome = sample(c("A", "B", "C"), 100, replace = TRUE), # categorical outcome group1 = sample(c("Male", "Female"), 100, replace = TRUE), # group variable 1 group2 = sample(c("Young", "Old"), 100, replace = TRUE) # group variable 2 ) # Summarize categorical outcome by two grouping variables cat_summary( yvar = "outcome", yname = c("A", "B", "C"), # ordered categories for outcome xvars = c("group1", "group2"), xname.list = list(c("Male", "Female"), c("Young", "Old")), data = data, yvar.display = "Outcome Category", xvars.display = c("Gender", "Age Group") )# Load a sample dataset data <- data.frame( outcome = sample(c("A", "B", "C"), 100, replace = TRUE), # categorical outcome group1 = sample(c("Male", "Female"), 100, replace = TRUE), # group variable 1 group2 = sample(c("Young", "Old"), 100, replace = TRUE) # group variable 2 ) # Summarize categorical outcome by two grouping variables cat_summary( yvar = "outcome", yname = c("A", "B", "C"), # ordered categories for outcome xvars = c("group1", "group2"), xname.list = list(c("Male", "Female"), c("Young", "Old")), data = data, yvar.display = "Outcome Category", xvars.display = c("Gender", "Age Group") )
Cumulative Distribution Function (CDF) plot for a biomarker.
cdf_plot(xvar, data, y.int = 5, xlim = NULL, xvar.display = xvar, group = NULL)cdf_plot(xvar, data, y.int = 5, xlim = NULL, xvar.display = xvar, group = NULL)
xvar |
The biomarker name. |
data |
The dataset. |
y.int |
Increasement interval on the y. |
xlim |
cdf plot range for xvar, when NULL, c(min(x), max(x)) will be used. |
xvar.display |
Display name of the biomarker. |
group |
A separate CDF line will be plotted for each group. |
A ggplot object representing the CDF inverse plot.
# Load a sample dataset data <- data.frame( biomarker = rnorm(100, mean = 50, sd = 10), group = sample(c("Group A", "Group B"), 100, replace = TRUE) ) # Basic CDF plot for a single biomarker without groups cdf_plot( xvar = "biomarker", data = data, y.int = 10, xlim = c(30, 70), xvar.display = "Biomarker Level" ) # CDF plot for a biomarker with groups cdf_plot( xvar = "biomarker", data = data, y.int = 10, xlim = c(30, 70), xvar.display = "Biomarker Level", group = "group" )# Load a sample dataset data <- data.frame( biomarker = rnorm(100, mean = 50, sd = 10), group = sample(c("Group A", "Group B"), 100, replace = TRUE) ) # Basic CDF plot for a single biomarker without groups cdf_plot( xvar = "biomarker", data = data, y.int = 10, xlim = c(30, 70), xvar.display = "Biomarker Level" ) # CDF plot for a biomarker with groups cdf_plot( xvar = "biomarker", data = data, y.int = 10, xlim = c(30, 70), xvar.display = "Biomarker Level", group = "group" )
This function evaluates the performance of a predictive model at a selected cutoff point.
cut_perf( yvar, censorvar = NULL, xvar, cutoff, dir, xvars.adj = NULL, data, type, yvar.display = yvar, xvar.display = xvar )cut_perf( yvar, censorvar = NULL, xvar, cutoff, dir, xvars.adj = NULL, data, type, yvar.display = yvar, xvar.display = xvar )
yvar |
Response variable name. |
censorvar |
Censoring variable name (0-censored, 1-event). |
xvar |
Biomarker name. |
cutoff |
Selected cutoff value. |
dir |
Direction for desired subgroup (">", ">=", "<", "<="). |
xvars.adj |
Other covariates to adjust when evaluating the performance. |
data |
Data frame containing the variables. |
type |
Type of analysis: "c" for continuous, "s" for survival, and "b" for binary. |
yvar.display |
Display name of response variable. |
xvar.display |
Display name of biomarker variable. |
A list containing various performance metrics and optionally, plots.
# Load a sample dataset data <- data.frame( survival_time = rexp(100, rate = 0.1), # survival time status = sample(c(0, 1), 100, replace = TRUE), # censoring status biomarker = rnorm(100, mean = 0, sd = 1), # biomarker levels covariate1 = rnorm(100, mean = 50, sd = 10) # an additional covariate ) # Perform cutoff performance evaluation for continuous outcome data$continuous_outcome <- rnorm(100, mean = 10, sd = 5) cut_perf( yvar = "continuous_outcome", xvar = "biomarker", cutoff = 0.5, dir = ">=", data = data, type = "c", yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level" ) # Perform cutoff performance evaluation for binary outcome data$binary_outcome <- sample(c(0, 1), 100, replace = TRUE) cut_perf( yvar = "binary_outcome", xvar = "biomarker", cutoff = 0, dir = "<=", data = data, type = "b", yvar.display = "Binary Outcome", xvar.display = "Biomarker Level" )# Load a sample dataset data <- data.frame( survival_time = rexp(100, rate = 0.1), # survival time status = sample(c(0, 1), 100, replace = TRUE), # censoring status biomarker = rnorm(100, mean = 0, sd = 1), # biomarker levels covariate1 = rnorm(100, mean = 50, sd = 10) # an additional covariate ) # Perform cutoff performance evaluation for continuous outcome data$continuous_outcome <- rnorm(100, mean = 10, sd = 5) cut_perf( yvar = "continuous_outcome", xvar = "biomarker", cutoff = 0.5, dir = ">=", data = data, type = "c", yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level" ) # Perform cutoff performance evaluation for binary outcome data$binary_outcome <- sample(c(0, 1), 100, replace = TRUE) cut_perf( yvar = "binary_outcome", xvar = "biomarker", cutoff = 0, dir = "<=", data = data, type = "b", yvar.display = "Binary Outcome", xvar.display = "Biomarker Level" )
Function for evaluating XGBoostSub_bin model performance.
eval_metric_bin(model, X_feature, y_label, pi, trt, Loss_type = "A_learning")eval_metric_bin(model, X_feature, y_label, pi, trt, Loss_type = "A_learning")
model |
The trained XGBoostSub_bin model object. |
X_feature |
The input features matrix. |
y_label |
The input y matrix. |
pi |
The propensity scores vector, which should range from 0 to 1, representing the probability of assignment to treatment. |
trt |
The treatment indicator vector. Should take values of 1 or -1, where 1 represents the treatment group and -1 represents the control group. |
Loss_type |
Type of loss function to use: "A_learning" or "Weight_learning". |
eval_metric: Function for Evaluating XGBoostSub_bin Model Performance
This function evaluates the performance of an XGBoostSub_bin model using a A-learning or weight-learning function.
Evaluation result of the XGBoostSub_bin model.
Function for evaluating XGBoostSub_con model performance.
eval_metric_con(model, X_feature, y_label, pi, trt, Loss_type = "A_learning")eval_metric_con(model, X_feature, y_label, pi, trt, Loss_type = "A_learning")
model |
The trained XGBoostSub_con model object. |
X_feature |
The input features matrix. |
y_label |
The input y matrix. |
pi |
The propensity scores vector, which should range from 0 to 1, representing the probability of assignment to treatment. |
trt |
The treatment indicator vector. Should take values of 1 or -1, where 1 represents the treatment group and -1 represents the control group. |
Loss_type |
Type of loss function to use: "A_learning" or "Weight_learning". |
eval_metric: Function for Evaluating XGBoostSub_con Model Performance
This function evaluates the performance of an XGBoostSub_con model using a A-learning or weight-learning function.
Evaluation result of the XGBoostSub_con model.
Function for evaluating XGBoostSub_sur model performance.
eval_metric_sur( model, X_feature, y_label, pi, trt, censor, Loss_type = "A_learning" )eval_metric_sur( model, X_feature, y_label, pi, trt, censor, Loss_type = "A_learning" )
model |
The trained XGBoostSub_sur model object. |
X_feature |
The input features matrix. |
y_label |
The input y matrix. |
pi |
The propensity scores vector, which should range from 0 to 1, representing the probability of assignment to treatment. |
trt |
The treatment indicator vector. Should take values of 1 or -1, where 1 represents the treatment group and -1 represents the control group. |
censor |
The censor status vector. Should take values of 1 or 0, where 1 represents censoring and 0 represents an observed event. |
Loss_type |
Type of loss function to use: "A_learning" or "Weight_learning". |
eval_metric: Function for Evaluating XGBoostSub_con Model Performance
This function evaluates the performance of an XGBoostSub_con model using a A-learning or weight-learning function.
Evaluation result of the XGBoostSub_sur model.
This function conducts fixed cutoff analysis for individual biomarker associated with binary outcome variables.
fixcut_bin( yvar, xvar, dir, cutoffs, data, method = "Fisher", yvar.display = yvar, xvar.display = xvar, vert.x = FALSE )fixcut_bin( yvar, xvar, dir, cutoffs, data, method = "Fisher", yvar.display = yvar, xvar.display = xvar, vert.x = FALSE )
yvar |
Binary response variable name. 0 represents controls and 1 represents cases. |
xvar |
Biomarker name. |
dir |
Cutoff direction for the desired subgroup. Options are ">", ">=", "<", or "<=". |
cutoffs |
A vector of candidate cutoffs. |
data |
The dataset containing the variables. |
method |
Method for cutoff selection. Options are "Fisher", "Youden", "Conc.Prob", "Accuracy", or "Kappa". - "Fisher": Minimizes the Fisher test p-value. - "Youden": Maximizes the Youden index. - "Conc.Prob": Maximizes sensitivity * specificity. - "Accuracy": Maximizes accuracy. - "Kappa": Maximizes Kappa coefficient. |
yvar.display |
Display name of the response variable. |
xvar.display |
Display name of the predictor variable. |
vert.x |
Whether to display the cutoff in a 90-degree angle when plotting (saves space). |
A list containing statistical summaries, selected cutoff statistics, selected cutoff value, confusion matrix, and a ggplot object for visualization.
# Load a sample dataset data <- data.frame( outcome = sample(c(0, 1), 100, replace = TRUE), biomarker = rnorm(100, mean = 0, sd = 1) ) # Perform fixed cutoff analysis using the "Fisher" method for a biomarker fixcut_bin( yvar = "outcome", xvar = "biomarker", dir = ">", cutoffs = seq(-2, 2, by = 0.5), data = data, method = "Fisher", yvar.display = "Binary Outcome", xvar.display = "Biomarker Level", vert.x = TRUE ) # Perform fixed cutoff analysis using the "Youden" method fixcut_bin( yvar = "outcome", xvar = "biomarker", dir = "<", cutoffs = seq(-2, 2, by = 0.5), data = data, method = "Youden", yvar.display = "Binary Outcome", xvar.display = "Biomarker Level", vert.x = FALSE ) # Perform fixed cutoff analysis using "Accuracy" method with different direction fixcut_bin( yvar = "outcome", xvar = "biomarker", dir = ">=", cutoffs = c(-1, 0, 1), data = data, method = "Accuracy", yvar.display = "Binary Outcome", xvar.display = "Biomarker Level", vert.x = TRUE )# Load a sample dataset data <- data.frame( outcome = sample(c(0, 1), 100, replace = TRUE), biomarker = rnorm(100, mean = 0, sd = 1) ) # Perform fixed cutoff analysis using the "Fisher" method for a biomarker fixcut_bin( yvar = "outcome", xvar = "biomarker", dir = ">", cutoffs = seq(-2, 2, by = 0.5), data = data, method = "Fisher", yvar.display = "Binary Outcome", xvar.display = "Biomarker Level", vert.x = TRUE ) # Perform fixed cutoff analysis using the "Youden" method fixcut_bin( yvar = "outcome", xvar = "biomarker", dir = "<", cutoffs = seq(-2, 2, by = 0.5), data = data, method = "Youden", yvar.display = "Binary Outcome", xvar.display = "Biomarker Level", vert.x = FALSE ) # Perform fixed cutoff analysis using "Accuracy" method with different direction fixcut_bin( yvar = "outcome", xvar = "biomarker", dir = ">=", cutoffs = c(-1, 0, 1), data = data, method = "Accuracy", yvar.display = "Binary Outcome", xvar.display = "Biomarker Level", vert.x = TRUE )
This function conducts fixed cutoff analysis for individual biomarker associated with continuous outcome variables.
fixcut_con( yvar, xvar, dir, cutoffs, data, method = "t.test", yvar.display = yvar, xvar.display = xvar, vert.x = FALSE )fixcut_con( yvar, xvar, dir, cutoffs, data, method = "t.test", yvar.display = yvar, xvar.display = xvar, vert.x = FALSE )
yvar |
Continuous response variable name. |
xvar |
Biomarker name. |
dir |
Cutoff direction for the desired subgroup. Options are ">", ">=", "<", or "<=". |
cutoffs |
A vector of candidate cutoffs. |
data |
The dataset containing the variables. |
method |
Method for cutoff selection. Currently only supports "t.test". - "t.test": Minimizes the t-test p-value. |
yvar.display |
Display name of the response variable. |
xvar.display |
Display name of the predictor variable. |
vert.x |
Whether to display the cutoff in a 90-degree angle when plotting (saves space). |
A list containing statistical summaries, selected cutoff statistics, selected cutoff value, group statistics, and a ggplot object for visualization.
# Load a sample dataset data <- data.frame( outcome = rnorm(100, mean = 10, sd = 5), biomarker = rnorm(100, mean = 0, sd = 1) ) # Perform fixed cutoff analysis using the "t.test" method with '>' direction fixcut_con( yvar = "outcome", xvar = "biomarker", dir = ">", cutoffs = seq(-2, 2, by = 0.5), data = data, method = "t.test", yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level", vert.x = TRUE ) # Perform fixed cutoff analysis with '<=' direction fixcut_con( yvar = "outcome", xvar = "biomarker", dir = "<=", cutoffs = c(-1, 0, 1), data = data, method = "t.test", yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level", vert.x = FALSE )# Load a sample dataset data <- data.frame( outcome = rnorm(100, mean = 10, sd = 5), biomarker = rnorm(100, mean = 0, sd = 1) ) # Perform fixed cutoff analysis using the "t.test" method with '>' direction fixcut_con( yvar = "outcome", xvar = "biomarker", dir = ">", cutoffs = seq(-2, 2, by = 0.5), data = data, method = "t.test", yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level", vert.x = TRUE ) # Perform fixed cutoff analysis with '<=' direction fixcut_con( yvar = "outcome", xvar = "biomarker", dir = "<=", cutoffs = c(-1, 0, 1), data = data, method = "t.test", yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level", vert.x = FALSE )
This function conducts fixed cutoff analysis for Individual Biomarker Associated with survival outcome variables.
fixcut_sur( yvar, censorvar, xvar, dir, cutoffs, data, method = "logrank", yvar.display = yvar, xvar.display = xvar, vert.x = FALSE )fixcut_sur( yvar, censorvar, xvar, dir, cutoffs, data, method = "logrank", yvar.display = yvar, xvar.display = xvar, vert.x = FALSE )
yvar |
Survival response variable name. |
censorvar |
Censoring variable. 0 indicates censored, 1 indicates an event. |
xvar |
Biomarker name. |
dir |
Cutoff direction for the desired subgroup. Options are ">", ">=", "<", or "<=". |
cutoffs |
A vector of candidate cutoffs. |
data |
The dataset containing the variables. |
method |
Method for cutoff selection. Currently only supports "logrank". - "logrank": Minimizes the logrank test p-value. |
yvar.display |
Display name of the response variable. |
xvar.display |
Display name of the predictor variable. |
vert.x |
Whether to display the cutoff in a 90-degree angle when plotting (saves space). |
A list containing statistical summaries, selected cutoff statistics, selected cutoff value, group statistics, and a ggplot object for visualization.
# Load a sample dataset data <- data.frame( time = rexp(100, rate = 0.1), # survival time status = sample(c(0, 1), 100, replace = TRUE), # censoring status biomarker = rnorm(100, mean = 0, sd = 1) # biomarker levels ) fixcut_sur( yvar = "time", censorvar = "status", xvar = "biomarker", dir = "<=", cutoffs = c(-1, 0, 1), data = data, method = "logrank", yvar.display = "Survival Time", xvar.display = "Biomarker Level", vert.x = FALSE )# Load a sample dataset data <- data.frame( time = rexp(100, rate = 0.1), # survival time status = sample(c(0, 1), 100, replace = TRUE), # censoring status biomarker = rnorm(100, mean = 0, sd = 1) # biomarker levels ) fixcut_sur( yvar = "time", censorvar = "status", xvar = "biomarker", dir = "<=", cutoffs = c(-1, 0, 1), data = data, method = "logrank", yvar.display = "Survival Time", xvar.display = "Biomarker Level", vert.x = FALSE )
Computes and plots the contrasts between treatment and control group based on a GAM for exploring the relationship be-tween treatment benefit and biomarker.
gam_ctr_plot( yvar, censorvar = NULL, xvar, xvars.adj = NULL, sxvars.adj = NULL, trtvar = NULL, type, data, k, title = "Group Contrast", ybreaks = NULL, xbreaks = NULL, rugcol.var = NULL, link.scale = TRUE, prt.sum = TRUE, prt.chk = FALSE, outlier.rm = FALSE )gam_ctr_plot( yvar, censorvar = NULL, xvar, xvars.adj = NULL, sxvars.adj = NULL, trtvar = NULL, type, data, k, title = "Group Contrast", ybreaks = NULL, xbreaks = NULL, rugcol.var = NULL, link.scale = TRUE, prt.sum = TRUE, prt.chk = FALSE, outlier.rm = FALSE )
yvar |
Response variable name. |
censorvar |
Censoring variable name (0-censored, 1-event). Required if type is "s" (survival). |
xvar |
Biomarker name. |
xvars.adj |
Potential confounding variables to adjust for using linear terms. |
sxvars.adj |
Potential confounding variables to adjust for using curves. |
trtvar |
Treatment variable that the contrast will build upon (treatment-control). |
type |
Type of response variable. Options are "c" for continuous, "s" for survival, and "b" for binary response. |
data |
The dataset containing the variables. |
k |
Upper limit on the degrees of freedom associated with an s smooth.When this k is too large, program will report error saying |
title |
Title of the plot. |
ybreaks |
Breaks on the y-axis. |
xbreaks |
Breaks on the x-axis. |
rugcol.var |
Variable name that defines the color of the rug. |
link.scale |
Whether to show the plot (y-axis) in the scale of the link function (linear predictor). |
prt.sum |
Whether to print summary or not. |
prt.chk |
Whether to print model diagnosis. |
outlier.rm |
Whether to remove outliers based on 1.5IQR. |
A list containing the p-value table, summarized p-value table, s-value table, summarized s-value table, and the plot.
# Load a sample dataset data <- data.frame( response = rnorm(100), biomarker = rnorm(100, mean = 50, sd = 10), censor = sample(c(0, 1), 100, replace = TRUE), treatment = sample(c(0, 1), 100, replace = TRUE), age = rnorm(100, mean = 60, sd = 10), group = sample(c("Group A", "Group B"), 100, replace = TRUE) ) # Generate a GAM contrast plot for a continuous response variable gam_ctr_plot( yvar = "response", xvar = "biomarker", trtvar = "treatment", type = "c", data = data, xvars.adj = "age", k = 5, title = "GAM Contrast Plot for Treatment vs. Control" ) # Generate a GAM contrast plot for survival analysis gam_ctr_plot( yvar = "response", censorvar = "censor", xvar = "biomarker", trtvar = "treatment", type = "s", data = data, k = 5, title = "GAM Contrast Plot for Survival Data" ) # Generate a GAM contrast plot for a binary response variable data$binary_response <- as.numeric(data$response > 0) gam_ctr_plot( yvar = "binary_response", xvar = "biomarker", trtvar = "treatment", type = "b", data = data, k = 5, title = "GAM Contrast Plot for Binary Outcome" )# Load a sample dataset data <- data.frame( response = rnorm(100), biomarker = rnorm(100, mean = 50, sd = 10), censor = sample(c(0, 1), 100, replace = TRUE), treatment = sample(c(0, 1), 100, replace = TRUE), age = rnorm(100, mean = 60, sd = 10), group = sample(c("Group A", "Group B"), 100, replace = TRUE) ) # Generate a GAM contrast plot for a continuous response variable gam_ctr_plot( yvar = "response", xvar = "biomarker", trtvar = "treatment", type = "c", data = data, xvars.adj = "age", k = 5, title = "GAM Contrast Plot for Treatment vs. Control" ) # Generate a GAM contrast plot for survival analysis gam_ctr_plot( yvar = "response", censorvar = "censor", xvar = "biomarker", trtvar = "treatment", type = "s", data = data, k = 5, title = "GAM Contrast Plot for Survival Data" ) # Generate a GAM contrast plot for a binary response variable data$binary_response <- as.numeric(data$response > 0) gam_ctr_plot( yvar = "binary_response", xvar = "biomarker", trtvar = "treatment", type = "b", data = data, k = 5, title = "GAM Contrast Plot for Binary Outcome" )
Generates a generalized additive model (GAM) plot for exploring the relationship between a response variable and a biomarker.
gam_plot( yvar, censorvar = NULL, xvar, xvars.adj = NULL, sxvars.adj = NULL, type, data, k, pred.type = "iterms", link.scale = TRUE, title = "Trend Plot", ybreaks = NULL, xbreaks = NULL, rugcol.var = NULL, add.points = FALSE, prt.sum = TRUE, prt.chk = FALSE, outlier.rm = FALSE, newdat = NULL )gam_plot( yvar, censorvar = NULL, xvar, xvars.adj = NULL, sxvars.adj = NULL, type, data, k, pred.type = "iterms", link.scale = TRUE, title = "Trend Plot", ybreaks = NULL, xbreaks = NULL, rugcol.var = NULL, add.points = FALSE, prt.sum = TRUE, prt.chk = FALSE, outlier.rm = FALSE, newdat = NULL )
yvar |
Response variable name. |
censorvar |
Censoring variable name for survival analysis (0-censored, 1-event). |
xvar |
Biomarker name. |
xvars.adj |
Potential confounding variables to adjust for using linear terms. |
sxvars.adj |
Potential confounding variables to adjust for using curve terms. |
type |
"c" for continuous, "s" for survival, and "b" for binary response. |
data |
The dataset containing the variables. |
k |
Upper limit on the degrees of freedom associated with an s smooth. |
pred.type |
"iterms" for trend of xvar, "response" for Y at the original scale. |
link.scale |
Whether to show the plot in the scale of the link function. |
title |
Title of the plot. |
ybreaks |
Breaks on the y-axis. |
xbreaks |
Breaks on the x-axis. |
rugcol.var |
Variable name defining the color of the rug and points. |
add.points |
Whether to add data points to the plot. |
prt.sum |
Whether to print summary or not. |
prt.chk |
Whether to print model diagnosis. |
outlier.rm |
Whether to remove outliers based on 1.5IQR. |
newdat |
User-supplied customized data for prediction and plotting. |
A list containing p-table, s-table, GAM summary, GAM check, and the plot.
# Load a sample dataset data <- data.frame( response = rnorm(100), biomarker = rnorm(100, mean = 50, sd = 10), censor = sample(c(0, 1), 100, replace = TRUE), age = rnorm(100, mean = 60, sd = 10), group = sample(c("Group A", "Group B"), 100, replace = TRUE) ) # Generate a GAM plot for a continuous response variable gam_plot( yvar = "response", xvar = "biomarker", type = "c", data = data, xvars.adj = "age", sxvars.adj = NULL, k = 5, pred.type = "iterms", title = "GAM Plot of Biomarker and Response" ) # Generate a GAM plot for survival analysis gam_plot( yvar = "response", censorvar = "censor", xvar = "biomarker", type = "s", data = data, k = 5, title = "GAM Survival Plot for Biomarker" ) # Generate a GAM plot for a binary response variable data$binary_response <- as.numeric(data$response > 0) gam_plot( yvar = "binary_response", xvar = "biomarker", type = "b", data = data, k = 5, pred.type = "response", title = "GAM Plot for Binary Response" )# Load a sample dataset data <- data.frame( response = rnorm(100), biomarker = rnorm(100, mean = 50, sd = 10), censor = sample(c(0, 1), 100, replace = TRUE), age = rnorm(100, mean = 60, sd = 10), group = sample(c("Group A", "Group B"), 100, replace = TRUE) ) # Generate a GAM plot for a continuous response variable gam_plot( yvar = "response", xvar = "biomarker", type = "c", data = data, xvars.adj = "age", sxvars.adj = NULL, k = 5, pred.type = "iterms", title = "GAM Plot of Biomarker and Response" ) # Generate a GAM plot for survival analysis gam_plot( yvar = "response", censorvar = "censor", xvar = "biomarker", type = "s", data = data, k = 5, title = "GAM Survival Plot for Biomarker" ) # Generate a GAM plot for a binary response variable data$binary_response <- as.numeric(data$response > 0) gam_plot( yvar = "binary_response", xvar = "biomarker", type = "b", data = data, k = 5, pred.type = "response", title = "GAM Plot for Binary Response" )
This function predicts the treatment assignment for each patient based on a cutoff value.
get_subgroup_results(model, X_feature, subgroup_label = NULL, cutoff = 0.5)get_subgroup_results(model, X_feature, subgroup_label = NULL, cutoff = 0.5)
model |
The trained XGBoost-based subgroup model. |
X_feature |
The data matrix containing patient features. |
subgroup_label |
(Optional) The subgroup labels. In real-world data, this information is typically unknown and only available in simulated data. If provided, the prediction accuracy will also be returned. |
cutoff |
The cutoff value for treatment assignment, defaulted to 0.5. |
A data frame containing each subject and assigned treatment (1 for treatment, 0 for control). If subgroup labels are provided, it also returns the prediction accuracy of the subgroup labels.
X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rnorm(100) # continuous outcome variable trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_con( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) subgroup_results=get_subgroup_results(model_A, X_data, subgroup_label=NULL, cutoff = 0.5)X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rnorm(100) # continuous outcome variable trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_con( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) subgroup_results=get_subgroup_results(model_A, X_data, subgroup_label=NULL, cutoff = 0.5)
This function calculates and plots the importance of biomarkers in a trained XGBoostSub_con, XGBoostSub_bin or XGBoostSub_sur model.
predictive_biomarker_imp(model)predictive_biomarker_imp(model)
model |
The trained XGBoost-based model. |
A barplot showing the biomarker importance.
X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rnorm(100) # continuous outcome variable trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_con( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) biomarker_imp=predictive_biomarker_imp(model_A)X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rnorm(100) # continuous outcome variable trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_con( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) biomarker_imp=predictive_biomarker_imp(model_A)
Computes the area under the receiver operating characteristic (ROC) curve for Biomarkers Associated with Binary Outcomes, and returns the results as a table.
roc_bin(yvar, xvars, dirs, data, yvar.display = yvar, xvars.display = xvars)roc_bin(yvar, xvars, dirs, data, yvar.display = yvar, xvars.display = xvars)
yvar |
Binary response variable name, where 0 represents controls and 1 represents cases. |
xvars |
A vector of biomarker names. |
dirs |
A vector of directions for the biomarkers. Options are "auto", ">", or "<". - "auto" (default): automatically determines in which group the median is higher and takes the direction accordingly. - ">": indicates that the biomarkers for the control group are higher than those for the case group (controls > t >= cases). - "<": indicates that the biomarkers for the control group are lower or equal to those for the case group (controls < t <= cases). |
data |
The dataset containing the variables. |
yvar.display |
Display name for the binary response variable. |
xvars.display |
Display names for the biomarkers. |
A table containing the AUC values for each biomarker.
# Load a sample dataset data <- data.frame( outcome = sample(c(0, 1), 100, replace = TRUE), biomarker1 = rnorm(100, mean = 0, sd = 1), biomarker2 = rnorm(100, mean = 5, sd = 2) ) # Compute AUC for a single biomarker with auto direction roc_bin( yvar = "outcome", xvars = "biomarker1", dirs = "auto", data = data, yvar.display = "Binary Outcome", xvars.display = "Biomarker 1" ) # Compute AUC for multiple biomarkers with specified directions roc_bin( yvar = "outcome", xvars = c("biomarker1", "biomarker2"), dirs = c("auto", "<"), data = data, yvar.display = "Binary Outcome", xvars.display = c("Biomarker 1", "Biomarker 2") )# Load a sample dataset data <- data.frame( outcome = sample(c(0, 1), 100, replace = TRUE), biomarker1 = rnorm(100, mean = 0, sd = 1), biomarker2 = rnorm(100, mean = 5, sd = 2) ) # Compute AUC for a single biomarker with auto direction roc_bin( yvar = "outcome", xvars = "biomarker1", dirs = "auto", data = data, yvar.display = "Binary Outcome", xvars.display = "Biomarker 1" ) # Compute AUC for multiple biomarkers with specified directions roc_bin( yvar = "outcome", xvars = c("biomarker1", "biomarker2"), dirs = c("auto", "<"), data = data, yvar.display = "Binary Outcome", xvars.display = c("Biomarker 1", "Biomarker 2") )
Generates ROC plots for different biomarkers associated with binary outcomes.
roc_bin_plot( yvar, xvars, dirs, data, yvar.display = yvar, xvars.display = xvars )roc_bin_plot( yvar, xvars, dirs, data, yvar.display = yvar, xvars.display = xvars )
yvar |
Binary response variable name, where 0 represents controls and 1 represents cases. |
xvars |
A vector of biomarker names. |
dirs |
A vector of directions for the biomarkers. Options are "auto", ">", or "<". - "auto" (default): automatically determines in which group the median is higher and takes the direction accordingly. - ">" indicates that the biomarkers for the control group are higher than those for the case group (controls > t >= cases). - "<" indicates that the biomarkers for the control group are lower or equal to those for the case group (controls < t <= cases). |
data |
The dataset containing the variables. |
yvar.display |
Display name for the binary response variable. |
xvars.display |
Display names for the biomarkers. |
ROC plots for different biomarkers associated with binary outcomes.
# Load a sample dataset data <- data.frame( outcome = sample(c(0, 1), 100, replace = TRUE), biomarker1 = rnorm(100, mean = 0, sd = 1), biomarker2 = rnorm(100, mean = 5, sd = 2) ) # Generate ROC plot for a single biomarker with auto direction roc_bin_plot( yvar = "outcome", xvars = "biomarker1", dirs = "auto", data = data, yvar.display = "Binary Outcome", xvars.display = "Biomarker 1" ) # Generate ROC plots for multiple biomarkers with specified directions roc_bin_plot( yvar = "outcome", xvars = c("biomarker1", "biomarker2"), dirs = c("auto", "<"), data = data, yvar.display = "Binary Outcome", xvars.display = c("Biomarker 1", "Biomarker 2") )# Load a sample dataset data <- data.frame( outcome = sample(c(0, 1), 100, replace = TRUE), biomarker1 = rnorm(100, mean = 0, sd = 1), biomarker2 = rnorm(100, mean = 5, sd = 2) ) # Generate ROC plot for a single biomarker with auto direction roc_bin_plot( yvar = "outcome", xvars = "biomarker1", dirs = "auto", data = data, yvar.display = "Binary Outcome", xvars.display = "Biomarker 1" ) # Generate ROC plots for multiple biomarkers with specified directions roc_bin_plot( yvar = "outcome", xvars = c("biomarker1", "biomarker2"), dirs = c("auto", "<"), data = data, yvar.display = "Binary Outcome", xvars.display = c("Biomarker 1", "Biomarker 2") )
Generates a scatter plot for exploring the relationship between a continuous response variable and a biomarker variable.
scat_cont_plot( yvar, xvar, data, ybreaks = NULL, xbreaks = NULL, yvar.display = yvar, xvar.display = xvar )scat_cont_plot( yvar, xvar, data, ybreaks = NULL, xbreaks = NULL, yvar.display = yvar, xvar.display = xvar )
yvar |
Continuous response variable name. |
xvar |
biomarker name. |
data |
The dataset containing the variables. |
ybreaks |
Breaks on the y-axis. |
xbreaks |
Breaks on the x-axis. |
yvar.display |
Display name for the response variable. |
xvar.display |
Display name for the biomarker variable. |
A list containing correlation coefficients, scatter plot, slope, and intercept.
data <- data.frame( outcome = rnorm(100, mean = 10, sd = 2), biomarker = rnorm(100, mean = 0, sd = 1) ) # Generate a scatter plot with default axis breaks scat_cont_plot( yvar = "outcome", xvar = "biomarker", data = data, yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level" ) # Generate a scatter plot with specified axis breaks scat_cont_plot( yvar = "outcome", xvar = "biomarker", data = data, ybreaks = seq(5, 15, by = 1), xbreaks = seq(-2, 2, by = 0.5), yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level" )data <- data.frame( outcome = rnorm(100, mean = 10, sd = 2), biomarker = rnorm(100, mean = 0, sd = 1) ) # Generate a scatter plot with default axis breaks scat_cont_plot( yvar = "outcome", xvar = "biomarker", data = data, yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level" ) # Generate a scatter plot with specified axis breaks scat_cont_plot( yvar = "outcome", xvar = "biomarker", data = data, ybreaks = seq(5, 15, by = 1), xbreaks = seq(-2, 2, by = 0.5), yvar.display = "Continuous Outcome", xvar.display = "Biomarker Level" )
This function evaluates subgroup performance based on different types of response variables.
subgrp_perf( yvar, censorvar = NULL, grpvar, grpname, xvars.adj = NULL, data, type, yvar.display = yvar, grpvar.display = grpvar )subgrp_perf( yvar, censorvar = NULL, grpvar, grpname, xvars.adj = NULL, data, type, yvar.display = yvar, grpvar.display = grpvar )
yvar |
The response variable name. |
censorvar |
(Optional) The censoring variable name (0-censored, 1-event). |
grpvar |
The subgroup variable name. |
grpname |
A vector of ordered subgroup names (values in the column of grpvar). |
xvars.adj |
(Optional) Other covariates to adjust when evaluating the performance. |
data |
The dataset containing the variables. |
type |
The type of response variable: "c" for continuous, "s" for survival, and "b" for binary. |
yvar.display |
Display name of the response variable. |
grpvar.display |
Display name of the group variable. |
A list containing subgroup performance results including logrank p-value, median and mean survival, Cox model p-value, ANOVA p-value, and more based on the specified response variable type.
# Load a sample dataset data <- data.frame( survival_time = rexp(100, rate = 0.1), # survival time status = sample(c(0, 1), 100, replace = TRUE), # censoring status group = sample(c("Low", "Medium", "High"), 100, replace = TRUE), # subgroup variable covariate = rnorm(100, mean = 50, sd = 10) # an additional covariate ) # Perform subgroup performance evaluation for survival analysis subgrp_perf( yvar = "survival_time", censorvar = "status", grpvar = "group", grpname = c("Low", "Medium", "High"), data = data, type = "s", yvar.display = "Survival Time", grpvar.display = "Risk Group" ) # Perform subgroup performance evaluation for continuous outcome data$continuous_outcome <- rnorm(100, mean = 10, sd = 5) subgrp_perf( yvar = "continuous_outcome", grpvar = "group", grpname = c("Low", "Medium", "High"), data = data, type = "c", yvar.display = "Continuous Outcome", grpvar.display = "Risk Group" ) # Perform subgroup performance evaluation for binary outcome data$binary_outcome <- sample(c(0, 1), 100, replace = TRUE) subgrp_perf( yvar = "binary_outcome", grpvar = "group", grpname = c("Low", "Medium", "High"), data = data, type = "b", yvar.display = "Binary Outcome", grpvar.display = "Risk Group" )# Load a sample dataset data <- data.frame( survival_time = rexp(100, rate = 0.1), # survival time status = sample(c(0, 1), 100, replace = TRUE), # censoring status group = sample(c("Low", "Medium", "High"), 100, replace = TRUE), # subgroup variable covariate = rnorm(100, mean = 50, sd = 10) # an additional covariate ) # Perform subgroup performance evaluation for survival analysis subgrp_perf( yvar = "survival_time", censorvar = "status", grpvar = "group", grpname = c("Low", "Medium", "High"), data = data, type = "s", yvar.display = "Survival Time", grpvar.display = "Risk Group" ) # Perform subgroup performance evaluation for continuous outcome data$continuous_outcome <- rnorm(100, mean = 10, sd = 5) subgrp_perf( yvar = "continuous_outcome", grpvar = "group", grpname = c("Low", "Medium", "High"), data = data, type = "c", yvar.display = "Continuous Outcome", grpvar.display = "Risk Group" ) # Perform subgroup performance evaluation for binary outcome data$binary_outcome <- sample(c(0, 1), 100, replace = TRUE) subgrp_perf( yvar = "binary_outcome", grpvar = "group", grpname = c("Low", "Medium", "High"), data = data, type = "b", yvar.display = "Binary Outcome", grpvar.display = "Risk Group" )
This function evaluates the performance of subgroups based on different types of response variables in predictive cases.
subgrp_perf_pred( yvar, censorvar = NULL, grpvar, grpname, trtvar, trtname, xvars.adj = NULL, data, type, yvar.display = yvar, grpvar.display = grpvar, trtvar.display = trtvar )subgrp_perf_pred( yvar, censorvar = NULL, grpvar, grpname, trtvar, trtname, xvars.adj = NULL, data, type, yvar.display = yvar, grpvar.display = grpvar, trtvar.display = trtvar )
yvar |
Response variable name. |
censorvar |
Censoring variable name (0-censored, 1-event). |
grpvar |
Subgroup variable name. |
grpname |
A vector of ordered subgroup names (values in the column of grpvar). |
trtvar |
Treatment variable name. |
trtname |
A vector of ordered treatment names (values in the column of trtvar). |
xvars.adj |
Other covariates to adjust when evaluating the performance. |
data |
The dataset. |
type |
"c" for continuous; "s" for "survival", and "b" for binary. |
yvar.display |
Display name of the response variable. |
grpvar.display |
Display name of the group variable. |
trtvar.display |
Display name of the treatment variable. |
A list containing the comparison results, group results, and possibly a plot.
# Load a sample dataset data <- data.frame( response = rnorm(100, mean = 10, sd = 5), # continuous response survival_time = rexp(100, rate = 0.1), # survival time status = sample(c(0, 1), 100, replace = TRUE), # censoring status group = sample(c("Low", "Medium", "High"), 100, replace = TRUE), # subgroup variable treatment = sample(c("A", "B"), 100, replace = TRUE) # treatment variable ) # Subgroup performance evaluation for predictive cases - survival analysis subgrp_perf_pred( yvar = "survival_time", censorvar = "status", grpvar = "group", grpname = c("Low", "Medium", "High"), trtvar = "treatment", trtname = c("A", "B"), data = data, type = "s", yvar.display = "Survival Time", grpvar.display = "Risk Group", trtvar.display = "Treatment" ) # Subgroup performance evaluation for predictive cases - continuous outcome subgrp_perf_pred( yvar = "response", grpvar = "group", grpname = c("Low", "Medium", "High"), trtvar = "treatment", trtname = c("A", "B"), data = data, type = "c", yvar.display = "Response", grpvar.display = "Risk Group", trtvar.display = "Treatment" ) # Subgroup performance evaluation for predictive cases - binary outcome data$binary_response <- sample(c(0, 1), 100, replace = TRUE) subgrp_perf_pred( yvar = "binary_response", grpvar = "group", grpname = c("Low", "Medium", "High"), trtvar = "treatment", trtname = c("A", "B"), data = data, type = "b", yvar.display = "Binary Response", grpvar.display = "Risk Group", trtvar.display = "Treatment" )# Load a sample dataset data <- data.frame( response = rnorm(100, mean = 10, sd = 5), # continuous response survival_time = rexp(100, rate = 0.1), # survival time status = sample(c(0, 1), 100, replace = TRUE), # censoring status group = sample(c("Low", "Medium", "High"), 100, replace = TRUE), # subgroup variable treatment = sample(c("A", "B"), 100, replace = TRUE) # treatment variable ) # Subgroup performance evaluation for predictive cases - survival analysis subgrp_perf_pred( yvar = "survival_time", censorvar = "status", grpvar = "group", grpname = c("Low", "Medium", "High"), trtvar = "treatment", trtname = c("A", "B"), data = data, type = "s", yvar.display = "Survival Time", grpvar.display = "Risk Group", trtvar.display = "Treatment" ) # Subgroup performance evaluation for predictive cases - continuous outcome subgrp_perf_pred( yvar = "response", grpvar = "group", grpname = c("Low", "Medium", "High"), trtvar = "treatment", trtname = c("A", "B"), data = data, type = "c", yvar.display = "Response", grpvar.display = "Risk Group", trtvar.display = "Treatment" ) # Subgroup performance evaluation for predictive cases - binary outcome data$binary_response <- sample(c(0, 1), 100, replace = TRUE) subgrp_perf_pred( yvar = "binary_response", grpvar = "group", grpname = c("Low", "Medium", "High"), trtvar = "treatment", trtname = c("A", "B"), data = data, type = "b", yvar.display = "Binary Response", grpvar.display = "Risk Group", trtvar.display = "Treatment" )
A dataset containing sample data for demonstrating the functionalities of the BioPred package.
data(tutorial_data)data(tutorial_data)
A data frame with the following columns:
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A biomarker variable.
Numeric. A continuous outcome variable.
Binary. A binary outcome variable, where 0 represents one class and 1 represents another class.
Numeric. The time in months, used for survival analysis.
Binary. Event indicator variable, where 0 indicates censoring and 1 indicates the event of interest occurred.
Binary. Ground truth of subgroup label. In real-world scenarios, this information is typically unavailable.
Binary. Treatment indicator variable, where 0 represents control and 1 represents treatment.
Factor. A categorical version of the treatment variable, with levels "Placebo" and "Treatment".
Factor.
This dataset is used to illustrate various functions within the BioPred package, including predictive modeling and subgroup analysis. The columns represent different types of data typically encountered in clinical studies.
data(tutorial_data) head(tutorial_data)data(tutorial_data) head(tutorial_data)
Function for training XGBoost model with customized loss function for binary outcomes
XGBoostSub_bin( X_data, y_data, trt, pi, Loss_type = "A_learning", params = list(), nrounds = 50, disable_default_eval_metric = 1, verbose = TRUE )XGBoostSub_bin( X_data, y_data, trt, pi, Loss_type = "A_learning", params = list(), nrounds = 50, disable_default_eval_metric = 1, verbose = TRUE )
X_data |
The input features matrix. |
y_data |
The input y matrix. |
trt |
The treatment indicator vector. Should take values of 1 or -1, where 1 represents the treatment group and -1 represents the control group. |
pi |
The propensity scores vector, which should range from 0 to 1, representing the probability of assignment to treatment. |
Loss_type |
Type of loss function to use: "A_learning" or "Weight_learning". |
params |
A list of additional parameters for the xgb.train function. |
nrounds |
Number of boosting rounds. Default is 50. |
disable_default_eval_metric |
If 1, default evaluation metric will be disabled. |
verbose |
Logical. If TRUE, training progress will be printed; if FALSE, no progress will be printed. |
XGBoostSub_bin: Function for Training XGBoost Model with Customized Loss Function for binary outcomes
This function trains an XGBoost model using a customized loss function based on the A-learning and weight-learning.
This function requires the 'xgboost' library. Make sure to install and load the 'xgboost' library before using this function.
After running this function, the returned model can be used like a regular xgboost model.
Trained XGBoostSub_bin model.
X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rbinom(100, 1, 0.5) # binary outcomes (0 or 1) trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_bin( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) # Train the model using Weight-learning loss model_W <- XGBoostSub_bin( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "Weight_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE )X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rbinom(100, 1, 0.5) # binary outcomes (0 or 1) trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_bin( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) # Train the model using Weight-learning loss model_W <- XGBoostSub_bin( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "Weight_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE )
Function for training XGBoost model with customized loss function for continuous outcomes
XGBoostSub_con( X_data, y_data, trt, pi, Loss_type = "A_learning", params = list(), nrounds = 50, disable_default_eval_metric = 1, verbose = TRUE )XGBoostSub_con( X_data, y_data, trt, pi, Loss_type = "A_learning", params = list(), nrounds = 50, disable_default_eval_metric = 1, verbose = TRUE )
X_data |
The input features matrix. |
y_data |
The input y matrix. |
trt |
The treatment indicator vector. Should take values of 1 or -1, where 1 represents the treatment group and -1 represents the control group. |
pi |
The propensity scores vector, which should range from 0 to 1, representing the probability of assignment to treatment. |
Loss_type |
Type of loss function to use: "A_learning" or "Weight_learning". |
params |
A list of additional parameters for the xgb.train function. |
nrounds |
Number of boosting rounds. Default is 50. |
disable_default_eval_metric |
If 1, default evaluation metric will be disabled. |
verbose |
Logical. If TRUE, training progress will be printed; if FALSE, no progress will be printed. |
XGBoostSub_con: Function for Training XGBoost Model with Customized Loss Function for continuous outcomes
This function trains an XGBoost model using a customized loss function based on the A-learning and weight-learning.
This function requires the 'xgboost' library. Make sure to install and load the 'xgboost' library before using this function.
After running this function, the returned model can be used like a regular xgboost model.
Trained XGBoostSub_con model.
X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rnorm(100) # continuous outcome variable trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_con( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) # Train the model using Weight-learning loss model_W <- XGBoostSub_con( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "Weight_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE )X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rnorm(100) # continuous outcome variable trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_con( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) # Train the model using Weight-learning loss model_W <- XGBoostSub_con( X_data = X_data, y_data = y_data, trt = trt, pi = pi, Loss_type = "Weight_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE )
Function for training XGBoost model with customized loss function for survival outcomes
XGBoostSub_sur( X_data, y_data, trt, pi, censor, Loss_type = "Weight_learning", params = list(), nrounds = 50, disable_default_eval_metric = 1, verbose = TRUE )XGBoostSub_sur( X_data, y_data, trt, pi, censor, Loss_type = "Weight_learning", params = list(), nrounds = 50, disable_default_eval_metric = 1, verbose = TRUE )
X_data |
The input features matrix. |
y_data |
The input y matrix. |
trt |
The treatment indicator vector. Should take values of 1 or -1, where 1 represents the treatment group and -1 represents the control group. |
pi |
The propensity scores vector, which should range from 0 to 1, representing the probability of assignment to treatment. |
censor |
The censor status vector. Should take values of 1 or 0, where 1 represents censoring and 0 represents an observed event. |
Loss_type |
Type of loss function to use: "A_learning" or "Weight_learning". |
params |
A list of additional parameters for the xgb.train function. |
nrounds |
Number of boosting rounds. Default is 50. |
disable_default_eval_metric |
If 1, default evaluation metric will be disabled. |
verbose |
Logical. If TRUE, training progress will be printed; if FALSE, no progress will be printed. |
XGBoostSub_sur: Function for Training XGBoost Model with Customized Loss Function for survival outcomes
This function trains an XGBoost model using a customized loss function based on the A-learning and weight-learning.
This function requires the 'xgboost' library. Make sure to install and load the 'xgboost' library before using this function.
Trained XGBoostSub_sur model.
X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rexp(100, rate = 0.1) # survival times, simulated as exponential trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 censor <- rbinom(100, 1, 0.7) # censoring indicator (1 = censored, 0 = observed) # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_sur( X_data = X_data, y_data = y_data, trt = trt, pi = pi, censor = censor, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) # Train the model using Weight-learning loss model_W <- XGBoostSub_sur( X_data = X_data, y_data = y_data, trt = trt, pi = pi, censor = censor, Loss_type = "Weight_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE )X_data <- matrix(rnorm(100 * 10), ncol = 10) # 100 samples with 10 features y_data <- rexp(100, rate = 0.1) # survival times, simulated as exponential trt <- sample(c(1, -1), 100, replace = TRUE) # treatment indicator (1 or -1) pi <- runif(100, min = 0.3, max = 0.7) # propensity scores between 0 and 1 censor <- rbinom(100, 1, 0.7) # censoring indicator (1 = censored, 0 = observed) # Define XGBoost parameters params <- list( max_depth = 3, eta = 0.1, subsample = 0.8, colsample_bytree = 0.8 ) # Train the model using A-learning loss model_A <- XGBoostSub_sur( X_data = X_data, y_data = y_data, trt = trt, pi = pi, censor = censor, Loss_type = "A_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE ) # Train the model using Weight-learning loss model_W <- XGBoostSub_sur( X_data = X_data, y_data = y_data, trt = trt, pi = pi, censor = censor, Loss_type = "Weight_learning", params = params, nrounds = 5, disable_default_eval_metric = 1, verbose = TRUE )