Doubly Robust Generalized Estimating Equations

Fit restricted mean models for the conditional association between an exposure and an outcome, given covariates. Three methods are implemented: O-estimation, where a nuisance model for the association between the covariates and the outcome is used; E-estimation where a nuisance model for the association between the covariates and the exposure is used, and doubly robust (DR) estimation where both nuisance models are used. In DR-estimation, the estimates will be consistent when at least one of the nuisance models is correctly specified, not necessarily both.


News

Changes in version 1.1.6 (2016-11-07)

o Added a helper function to obtain score residuals from clogit given the estimated coefficients, the observed outcomes, the design matrix and the id vector.

o Fixed a bug when using retrospective conditional logistic regression

o Add doubly robust estimation in conditional logistic models

Changes in version 1.1.5 (2016-04-25)

o Bugfixes

Changes in version 1.1.4 (2016-01-13)

o In some data manipulation steps, the data.table package is used to obtain better speed.

o For estimation in conditional logistic models only outcome-discordant clusters are used

o Calculation of residuals for estimating equations for conditional logistic models implemented in C++ using the Rcpp and RcppArmadillo packages.

o Drops unused levels for factor variables

Changes in version 1.1.3 (2015-06-23)

o Fixed a bug in 'drgeeData' which caused an error when one of the nuisance models only contained an intercept

Changes in version 1.1.2 (2015-05-18)

o Fixed a bug in 'drgeeData' for the 'estimation.method' which happened when 'cond = TRUE' and 'olink = "logit"'.

Changes in version 1.1.1 (2015-05-09)

o Fixed a bug in 'drgeeData' that caused error when no data argument is supplied.

o There was an error in the previous version, when the function 'drgeeData' removed outcome concordant cluster for conditional methods. This is now corrected such that this only happens for conditional logistic methods.

Changes in version 1.1.0 (2015-04-20)

o The arguments 'outcome', 'exposure' and 'clusterid' can be supplied as vectors or as a string.

o Changed the interface for the 'drgee' function. Now the argument 'estimationMethod' has to be "dr","o", or "e" corresponding to previous choices "dr", "obe" and "ebe". When 'estimationMethod="o"', the user can supply an argument 'exposure' instead of 'emodel'. Similarly, when 'estimationMethod="e"', the user can supply an argument 'outcome' instead of 'omodel'.

o Changed the functions 'drgee' and 'gee' such that they always return the variance as a matrix.

o Added a function 'gee' which calculates all coefficients in a regression.

o Added an option 'cond' for 'drgee' and 'gee' for models with cluster-specific intercepts.

Changes in version 1.0.1 (2014-01-27)

o Fixed a bug in 'drFit'. The calculation of the Jacobian was incorrect for doubly robust estimation with outcome link logit, resulting in inconsistent estimates of standard errors. This bug is now fixed.

Changes in Version 1.0 (2013-12-19)

o First version released on CRAN

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("drgee")

1.1.6 by Johan Zetterqvist, 10 months ago


Browse source code at https://github.com/cran/drgee


Authors: Johan Zetterqvist <johan.zetterqvist@ki.se> , Arvid Sjölander <arvid.sjolander@ki.se> with contributions from Alexander Ploner.


Documentation:   PDF Manual  


Task views: Robust Statistical Methods


GPL-2 | GPL-3 license


Depends on stats, nleqslv, survival, Rcpp, data.table

Linking to Rcpp, RcppArmadillo


Depended on by AF.


See at CRAN