sscaROC {caROC}

Covariate-adjusted continuous biomarker evaluations for specific population.

Description

Provides evalution for continuous biomarkers at controlled sensitivity/specificity level, or ROC curve in specified sub-population.

Usage

sscaROC(diseaseData, controlData, userFormula, target_covariates,
control_sensitivity = NULL, control_specificity = NULL, mono_resp_method = "ROC",
whichSE = "sample", global_ROC_controlled_by = "sensitivity", nbootstrap = 100,
CI_alpha = 0.95, logit_CI = TRUE, verbose = TRUE)

Arguments

diseaseData

Data from patients including dependent (biomarker) and independent (covariates) variables.

controlData

Data from controls including dependent (biomarker) and independent (covariates) variables.

userFormula

A character string to represent the function for covariate adjustment. For example, let Y denote biomarker, Z1 and Z2 denote two covariates. Then userFormula = "Y ~ Z1 + Z2".

target_covariates

Covariates of the interested sub-population. It could be a vector, e.g. c(1, 0.5, 0.8), or a matrix, e.g. target_covariates = matrix(c(1, 0.7, 0.9, 1, 0.8, 0.8), 2, 3, byrow = TRUE)

control_sensitivity

The level(s) of sensitivity to be controlled at. Could be a scalar (e.g. 0.7) or a numeric vector (e.g. c(0.7, 0.8, 0.9)).

control_specificity

The level(s) of specificity to be controlled at. Could be a scalar (e.g. 0.7) or a numeric vector (e.g. c(0.7, 0.8, 0.9)).

mono_resp_method

The method used to restore monotonicity of the ROC curve or computed sensitivity/specificity value. It should one from the following: "none", "ROC". "none" is not applying any monotonicity respecting method. "ROC" is to apply ROC-based monotonicity respecting approach. Default value is "ROC".

whichSE

The method used to compute standard error. It should be one from the following: "sample", "bootstrap", meaning to calculate the standard error using sample-based approach or bootstrap. Default is "sample".

global_ROC_controlled_by

Whether sensitivity/specificity is used to control when computing global ROC. It should one from the following: "sensitivity", "specificity". Default is "sensitivity".

nbootstrap

Number of boostrap iterations. Default is 100.

CI_alpha

Percentage of confidence interval. Default is 0.95.

logit_CI

Whether to apply logit-based confidence interval. Logit-transformed CI has been identified to be more robust near border area.

verbose

Whether to print out messages. Default value is true.

Value

If control_sensitivity or control_specificity is provided, compute covariate-adjusted specificity (sensitivity) at controlled sensitivity (specificity) level.

Estimate

Covariate-adjusted sensitivity/specificity.

SE

Estimated standard error.

CI

Estimated confidence intervals.

If both control_sensitivity and control_specificity are null, compuate covariate-adjusted ROC curve.

sensitivity

Estimated sensitivity.

specificity

Estimated specificity.

mono_adj

Monotonicity adjustment method.

Author(s)

Ziyi.li <[e-mail]>

Examples

n1 = n0 = 1000
## generate data
Z_D1 <- rbinom(n1, size = 1, prob = 0.3)
Z_D2 <- rnorm(n1, 0.8, 1)
Z_C1 <- rbinom(n0, size = 1, prob = 0.7)
Z_C2 <- rnorm(n0, 0.8, 1)
Y_C_Z0 <- rnorm(n0, 0.1, 1)
Y_D_Z0 <- rnorm(n1, 1.1, 1)
Y_C_Z1 <- rnorm(n0, 0.2, 1)
Y_D_Z1 <- rnorm(n1, 0.9, 1)
M0 <- Y_C_Z0 * (Z_C1 == 0) + Y_C_Z1 * (Z_C1 == 1) + Z_C2
M1 <- Y_D_Z0 * (Z_D1 == 0) + Y_D_Z1 * (Z_D1 == 1) + 1.5 * Z_D2
diseaseData <- data.frame(M = M1, Z1 = Z_D1, Z2 = Z_D2)
controlData <- data.frame(M = M0, Z1 = Z_C1, Z2 = Z_C2)
userFormula = "M~Z1+Z2"
target_covariates = c(1, 0.7, 0.9)
res <- sscaROC(diseaseData,controlData,
               userFormula = userFormula,
               control_sensitivity = c(0.2,0.8, 0.9),
               target_covariates = target_covariates,
               control_specificity = NULL,
               mono_resp_method = "none",
               whichSE = "sample",nbootstrap = 100,
               CI_alpha = 0.95, logit_CI = TRUE)


## bootstrap-based variance estimation
res <- sscaROC(diseaseData,controlData,
               userFormula = userFormula,
               control_sensitivity = c(0.2,0.8, 0.9),
               target_covariates = target_covariates,
               control_specificity = NULL,
               mono_resp_method = "none",
               whichSE = "bootstrap",nbootstrap = 100,
               CI_alpha = 0.95, logit_CI = TRUE)
## monotonization by ROC-based
res <- sscaROC(diseaseData,controlData,
               userFormula = userFormula,
               control_sensitivity = c(0.2,0.8, 0.9),
               target_covariates = target_covariates,
               control_specificity = NULL,
               mono_resp_method = "ROC",
               whichSE = "bootstrap",nbootstrap = 100,
               CI_alpha = 0.95, logit_CI = TRUE)
## control specificity
res <- sscaROC(diseaseData,controlData,
               userFormula = userFormula,
               control_sensitivity = NULL,
               target_covariates = target_covariates,
               control_specificity = c(0.2,0.8, 0.9),
               mono_resp_method = "ROC",
               whichSE = "bootstrap",nbootstrap = 100,
               CI_alpha = 0.95, logit_CI = TRUE)
### get ROC curves
myROC <- sscaROC(diseaseData,
                 controlData,
                 userFormula,
                 target_covariates,
                 global_ROC_controlled_by = "sensitivity",
                 mono_resp_method = "none")