mele: function to compute the maximum estimated likelihood...

View source: R/mele.R

meleR Documentation

function to compute the maximum estimated likelihood estimator

Description

This function computes the maximum estimated likelihood estimator (MELE) of regression parameters, which assess treatment-biomarker interactions in studies with two-phase sampling in randomized clinical trials. The function has an option to incorporate the independence between a randomized treatment and the baseline markers.

Usage

mele(data, response, treatment, BaselineMarker, extra = NULL, phase, 
ind = TRUE, maxit=2000)

Arguments

data

A data frame used to access the following data. Each row contains the response and predictors of a study participant. All variables are numerical.

response

A character string of column name, corresponds to one column of the data frame, which is used to store a numeric vector of response. The response variable should be coded as 1 for cases and 0 for controls.

treatment

A character string of column name, corresponds to one column of the data frame, which is used to store a binary vector of the treatment . The treatment variable should be coded as 1 for treatment and 0 for placebo.

BaselineMarker

A character string of column name, corresponds to one column of the data frame, which is used to store a vector of biomarker that is assessed for interaction with the treatment. The BaselineMarker variable is missing for those who are not sampled in the second phase.

extra

A string vector of column name(s), corresponds to one or more column(s) of the data frame, which are used to store the extra covariate(s) to be adjusted for in addition to treatment and biomarker. All extra variables are missing for those who are not sampled in the second phase.

phase

A character string of column name, correspond to one column of the data frame, which is used to store the indicator of two-phase sampling (1: not being sampled for measuring biomarker; 2: being sampled for measuring biomarker).

ind

A logical flag. TRUE indicates incorporating the independence between the randomized treatment and the baseline markers.

maxit

A integer number of the maximal number of iteration.

Details

The function returns estimates, standard errors, and p values for MELE of a regression model for treatment-biomarker interaction studies with two-phase sampling in randomized trials, response ~ treatment + biomarker + treatment*biomarker + other covariates. Treatment and response are available for all the samples, while baseline biomarker data are available for a subset of samples. The mele can incorporate the independence between the treatment and baseline biomarkers ascertained in the phase-two sample.

Value

beta

Estimated parameter

stder

Standard error

pVal

p value

Author(s)

James Y. Dai

References

J. Y. Dai, M. LeBlanc, and C. Kooperberg. Semiparametric estimation exploiting covariate independence in two-phase randomized trials. Biometrics, 65(1):178-187, 2009.

See Also

spmle

Examples

## Load the example data
data(whiBioMarker)
## Here is an example of MELE with exploiting independent and with confounding factors:

melIndExtra <- mele(data=whiBioMarker,  ## dataset
          response="stroke",	## response variable
          treatment="hrtdisp",		## treatment variable
          BaselineMarker="papbl",		## environment variable
          extra=c(
             "age" 	## age
                		## physical activity levels
              , "dias" 	## diabetes
              , "hyp" ## hypertension
              , "syst" 	## systolic
              , "diabtrt"	## diastolic BP
              , "lmsepi" ## waist:hip ratio
              ),	## extra variable(s)
          phase="phase",	## phase indicator
          ind=TRUE	## independent or non-indepentent
)

TwoPhaseInd documentation built on March 18, 2022, 7:52 p.m.