Package 'ROCSI' reference manual

Title:	Receiver Operating Characteristic Based Signature Identification
Description:	Optimal linear combination predictive signatures for maximizing the area between two Receiver Operating Characteristic (ROC) curves (treatment vs. control).
Authors:	Xin Huang [aut, cre, cph]
Maintainer:	Xin Huang <[email protected]>
License:	GPL (>= 2)
Version:	0.1.0
Built:	2024-12-04 07:07:03 UTC
Source:	CRAN

AUC

Description

Empirical AUC estimate

Usage

AUC(outcome, predict)
AUC(outcome, predict)

Arguments

`outcome`	binary outcome (1: desired outcome; 0: otherwise)
`predict`	prediction score

Details

Function for AUC when input is X and Y.

Value

a numeric value of empirical estimation of area under the ROC curves

Examples

# no run
# no run

beta2theta

Description

Function to translate beta into theta, the n-sphere constrain

Usage

beta2theta(beta)
beta2theta(beta)

Arguments

beta

estimates of coefficient beta

Details

Function to translate beta into theta, the n-sphere constrain

Value

a numeric vector for theta (dimension-1)

Examples

# no run
# no run

C.index

Description

Empirical c-index estimate

Usage

C.index(yvar, score, censorvar, data)
C.index(yvar, score, censorvar, data)

Arguments

`yvar`	column name for observed time
`score`	column name for marker value
`censorvar`	column name for censor (1 is event, 0 is censored)
`data`	input data matrix

Details

Function for c-index when input is X and Y.

Value

a numeric value of empirical estimation of c-index

Examples

# no run
# no run

cvfolds0

Description

internal function for generating CV fold index

Usage

cvfolds0(X, Y, idx, nfolds = 5)
cvfolds0(X, Y, idx, nfolds = 5)

Arguments

`X`	marker matrix for non-responders
`Y`	marker matrix for responders
`idx`	m*n by 2 matrix for row index of marker matrix, first column is row index in X; second column is for Y
`nfolds`	the cross-validation folds

Details

Function for generate CV fold index

Value

a vector containing CV fold index for each row in Z

Examples

# no run
# no run

data.gen

Description

Function for simulated data generation

Usage

data.gen(
  n,
  k,
  prevalence = sqrt(0.5),
  prog.eff = 1,
  sig2,
  y.sig2,
  rho,
  rhos.bt.real,
  a.constent
)
data.gen(
  n,
  k,
  prevalence = sqrt(0.5),
  prog.eff = 1,
  sig2,
  y.sig2,
  rho,
  rhos.bt.real,
  a.constent
)

Arguments

`n`	Total sample size
`k`	Number of markers
`prevalence`	prevalence of predictive biomarkers with values above the cutoff
`prog.eff`	effect size $beta$ for prognostic biomarker
`sig2`	standard deviation of each marker
`y.sig2`	Standard Deviation of the error term in the linear component
`rho`	rho*sig2 is the entries for covariance matrix between pairs of different k markers
`rhos.bt.real`	correlation between each prognostic and predictive markers
`a.constent`	a constant is set such that there is no overall treatment effect

Details

Function for simulated data generation

Value

A list of simulated clinical trial data with heterogeneous prognostic and predictive biomarkers

Examples

n <- 500
k <- 10
prevalence <- sqrt(0.5)
rho<-0.2
sig2 <- 2
rhos.bt.real <- c(0, rep(0.1, (k-3)))*sig2
y.sig2 <- 1
prog.eff <- 0.5
effect.size <- 1
a.constent <- effect.size/(2*(1-prevalence))
ObsData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                    sig2=sig2, y.sig2=y.sig2, rho=rho,
                    rhos.bt.real=rhos.bt.real, a.constent=a.constent)
n <- 500
k <- 10
prevalence <- sqrt(0.5)
rho<-0.2
sig2 <- 2
rhos.bt.real <- c(0, rep(0.1, (k-3)))*sig2
y.sig2 <- 1
prog.eff <- 0.5
effect.size <- 1
a.constent <- effect.size/(2*(1-prevalence))
ObsData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                    sig2=sig2, y.sig2=y.sig2, rho=rho,
                    rhos.bt.real=rhos.bt.real, a.constent=a.constent)

grad.sub

Description

Internal function of grad_square in the GCV

Usage

grad.sub(z, beta)
grad.sub(z, beta)

Arguments

`z`	(m x n) x p data matrix as prepared for ROCSI
`beta`	estimates of coefficient beta

Details

Internal function of grad_square in the GCV

Value

grad_square in the GCV

Examples

# no run
# no run

gradsqr

Description

Internal function for HIC calculation

Usage

gradsqr(beta, Z0, index, w = 1)
gradsqr(beta, Z0, index, w = 1)

Arguments

`beta`	estimates of coefficient beta
`Z0`	(m x n) x p Z matrix as prepared for ROCSI
`index`	m*n by 2 matrix for the subindex for the pair difference in Z
`w`	a vector of weights Z (can be used for inverse probability weighting for missing data, default is 1)

Details

Internal function for HIC calculation

Value

gradient square for the GCV.

Examples

# no run
# no run

hessAUC

Description

function for Hessian matrix of AUC

Usage

hessAUC(beta, Z, w = 1)
hessAUC(beta, Z, w = 1)

Arguments

`beta`	estimates of coefficient beta
`Z`	(m x n) x p data matrix as prepared for ROCSI
`w`	a vector of weights Z (can be used for inverse probability weighting for missing data, default is 1)

Details

function for Hessian matrix of AUC

Value

Hessian matrix of AUC.

Examples

# no run
# no run

hessAUC.sub

Description

Internal function for hessAUC

Usage

hessAUC.sub(z, beta)
hessAUC.sub(z, beta)

Arguments

`z`	(m x n) x p data matrix as prepared for ROCSI
`beta`	estimates of coefficient beta

Details

Internal function for hessAUC

Value

Hessian matrix components.

Examples

# no run
# no run

HIC

Description

function for HIC calculation

Usage

HIC(beta, Z, index, w = 1)
HIC(beta, Z, index, w = 1)

Arguments

`beta`	estimates of coefficient beta
`Z`	matrix prepared for ROCSI
`index`	m*n by 2 matrix for the subindex for the pair difference in Z
`w`	a vector of weights Z (can be used for inverse probability weighting for missing data, default is 1)

Details

Function for HIC calculation

Value

A numeric value with corresponding HIC

Examples

# no run
# no run

MClogit

Description

function for modified covariate methods based on glmnet

Usage

MClogit(
  dataset,
  yvar,
  xvars,
  trtvar,
  cvar = NULL,
  nfolds = 5,
  type = "binary",
  newx = NULL,
  bestsub = "lambda.1se",
  type.measure = "auc"
)
MClogit(
  dataset,
  yvar,
  xvars,
  trtvar,
  cvar = NULL,
  nfolds = 5,
  type = "binary",
  newx = NULL,
  bestsub = "lambda.1se",
  type.measure = "auc"
)

Arguments

`dataset`	data matrix for training dataset
`yvar`	column name for outcome
`xvars`	a string vector of column names for input markers
`trtvar`	column name for treatment (the column should contain binary code with 1 being treatment and 0 being control)
`cvar`	column name for censor (the column should contain binary code with 1 being event and 0 being censored)
`nfolds`	n fold CV used for cv.glmnet
`type`	outcome type ("binary" for binary outcome and "survival" for time-to-event outcome)
`newx`	data matrix for testing dataset X
`bestsub`	criteria for best lambda, used by glmnet
`type.measure`	type of measure used by glmnet

Details

function for ROCSI

Value

A list with ROCSI output

x.logit: final beta estimated from MClogit
predScore: a data.frame of testing data and its predictive signature scores (based on beta.aABC) for each subjects
abc: ABC in testing dataset based on optimal beta
fit.cv: the fitted glmnet object

Examples

n <- 100
k <- 5
prevalence <- sqrt(0.5)
rho<-0.2
sig2 <- 2
rhos.bt.real <- c(0, rep(0.1, (k-3)))*sig2
y.sig2 <- 1
yvar="y.binary"
xvars=paste("x", c(1:k), sep="")
trtvar="treatment"
prog.eff <- 0.5
effect.size <- 1
a.constent <- effect.size/(2*(1-prevalence))
ObsData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                    sig2=sig2, y.sig2=y.sig2, rho=rho,
                    rhos.bt.real=rhos.bt.real, a.constent=a.constent)
TestData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                     sig2=sig2, y.sig2=y.sig2, rho=rho,
                     rhos.bt.real=rhos.bt.real, a.constent=a.constent)
bst.mod <- MClogit(dataset=ObsData$data, yvar=yvar, xvars=xvars,
trtvar=trtvar, nfolds = 5, newx=TestData$data,
type="binary", bestsub="lambda.1se")
bst.mod$abc
bst.mod$x.logit[-1,1]
n <- 100
k <- 5
prevalence <- sqrt(0.5)
rho<-0.2
sig2 <- 2
rhos.bt.real <- c(0, rep(0.1, (k-3)))*sig2
y.sig2 <- 1
yvar="y.binary"
xvars=paste("x", c(1:k), sep="")
trtvar="treatment"
prog.eff <- 0.5
effect.size <- 1
a.constent <- effect.size/(2*(1-prevalence))
ObsData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                    sig2=sig2, y.sig2=y.sig2, rho=rho,
                    rhos.bt.real=rhos.bt.real, a.constent=a.constent)
TestData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                     sig2=sig2, y.sig2=y.sig2, rho=rho,
                     rhos.bt.real=rhos.bt.real, a.constent=a.constent)
bst.mod <- MClogit(dataset=ObsData$data, yvar=yvar, xvars=xvars,
trtvar=trtvar, nfolds = 5, newx=TestData$data,
type="binary", bestsub="lambda.1se")
bst.mod$abc
bst.mod$x.logit[-1,1]

pair.diff

Description

internal function for generating Z matrix (binary endpoint)

Usage

pair.diff(X, Y, A)
pair.diff(X, Y, A)

Arguments

`X`	marker matrix for non-responders
`Y`	marker matrix for responders
`A`	Treatment arm indicator (1 is treatment, 0 is control)

Details

Function for generate Z matrix for binary endpoint

Value

A list of prepared data input for ROCSI

Examples

# no run
# no run

pair.diff.surv

Description

internal function for generating Z matrix (time-to-event endpoint)

Usage

pair.diff.surv(X, Y, A, C)
pair.diff.surv(X, Y, A, C)

Arguments

`X`	marker matrix
`Y`	a vector for observed time
`A`	a vector for Treatment arm indicator (1 is treatment, 0 is control)
`C`	a vector for censor (1 is event, 0 is censored)

Details

Function for generate Z matrix for time-to-event endpoint

Value

A list of prepared data input for ROCSI

Examples

# no run
# no run

ROCSI

Description

function for ROCSI

Usage

ROCSI(
  Dtrain,
  Dtest = NULL,
  yvar,
  xvars,
  trtvar,
  cvar = NULL,
  nfolds = 5,
  type = "binary"
)
ROCSI(
  Dtrain,
  Dtest = NULL,
  yvar,
  xvars,
  trtvar,
  cvar = NULL,
  nfolds = 5,
  type = "binary"
)

Arguments

`Dtrain`	data matrix for training dataset
`Dtest`	optional data matrix for testing dataset
`yvar`	column name for outcome
`xvars`	a string vector of column names for input markers
`trtvar`	column name for treatment (the column should contain binary code with 1 being treatment and 0 being control)
`cvar`	column name for censor (the column should contain binary code with 1 being event and 0 being censored)
`nfolds`	n fold CV used for cv.glmnet
`type`	outcome type ("binary" for binary outcome and "survival" for time-to-event outcome)

Details

function for ROCSI

Value

A list with ROCSI output

beta.aABC: final beta estimated from ROCSI based on $ABC^{(acv)}$
beta.1se: final beta estimated from lambda.1se based on nfold CV
lambda.aABC: optimal lambda selected by optimizing $ABC^{(acv)}$
fit.cv: fitted cv.glmnet model
log: log matrix of all lambdas and ABCs
abc.test: ABC in testing dataset based on optimal beta
abc.test1se: ABC in testing dataset based on 1se beta
predScore: a data.frame of testing data and its predictive signature scores (based on beta.aABC) for each subjects
predScore.1se: a data.frame of testing data and its predictive signature scores (based on beta.1se) for each subjects

Examples

n <- 100
k <- 5
prevalence <- sqrt(0.5)
rho<-0.2
sig2 <- 2
rhos.bt.real <- c(0, rep(0.1, (k-3)))*sig2
y.sig2 <- 1
yvar="y.binary"
xvars=paste("x", c(1:k), sep="")
trtvar="treatment"
prog.eff <- 0.5
effect.size <- 1
a.constent <- effect.size/(2*(1-prevalence))
ObsData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                    sig2=sig2, y.sig2=y.sig2, rho=rho,
                    rhos.bt.real=rhos.bt.real, a.constent=a.constent)
TestData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                     sig2=sig2, y.sig2=y.sig2, rho=rho,
                     rhos.bt.real=rhos.bt.real, a.constent=a.constent)
bst.aabc <- ROCSI(Dtrain=ObsData$data, Dtest = TestData$data, yvar=yvar,
xvars=xvars, trtvar=trtvar, cvar=NULL, nfolds=5, type="binary")
bst.aabc$beta.aABC
bst.aabc$log
bst.aabc$abc.test
bst.aabc$beta.1se
bst.aabc$abc.test1se
n <- 100
k <- 5
prevalence <- sqrt(0.5)
rho<-0.2
sig2 <- 2
rhos.bt.real <- c(0, rep(0.1, (k-3)))*sig2
y.sig2 <- 1
yvar="y.binary"
xvars=paste("x", c(1:k), sep="")
trtvar="treatment"
prog.eff <- 0.5
effect.size <- 1
a.constent <- effect.size/(2*(1-prevalence))
ObsData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                    sig2=sig2, y.sig2=y.sig2, rho=rho,
                    rhos.bt.real=rhos.bt.real, a.constent=a.constent)
TestData <- data.gen(n=n, k=k, prevalence=prevalence, prog.eff=prog.eff,
                     sig2=sig2, y.sig2=y.sig2, rho=rho,
                     rhos.bt.real=rhos.bt.real, a.constent=a.constent)
bst.aabc <- ROCSI(Dtrain=ObsData$data, Dtest = TestData$data, yvar=yvar,
xvars=xvars, trtvar=trtvar, cvar=NULL, nfolds=5, type="binary")
bst.aabc$beta.aABC
bst.aabc$log
bst.aabc$abc.test
bst.aabc$beta.1se
bst.aabc$abc.test1se

theta2beta

Description

Function to translate theta into beta

Usage

theta2beta(theta)
theta2beta(theta)

Arguments

theta

n-sphere coordination

Details

Function to translate beta into theta, the n-sphere constrain

Value

a numeric vector for beta (dimension+1)

Examples

# no run
# no run

Package 'ROCSI'

Help Index

AUC

Description

Usage

Arguments

Details

Value

Examples

beta2theta

Description

Usage

Arguments

Details

Value

Examples

C.index

Description

Usage

Arguments

Details

Value

Examples

cvfolds0

Description

Usage

Arguments

Details

Value

Examples

data.gen

Description

Usage

Arguments

Details

Value

Examples

grad.sub

Description

Usage

Arguments

Details

Value

Examples

gradsqr

Description

Usage

Arguments

Details

Value

Examples

hessAUC

Description

Usage

Arguments

Details

Value

Examples

hessAUC.sub

Description

Usage

Arguments

Details

Value

Examples

HIC

Description

Usage

Arguments

Details

Value

Examples

MClogit

Description

Usage

Arguments

Details

Value

Examples

pair.diff