The geess function analyzes small-sample clustered or longitudinal data using modified generalized estimating equations (GEE) with bias-adjusted covariance estimator. This is an extension of geessbin function (Ishii et al., 2024). The geess function provides any combination of three GEE methods (conventional and two modified GEE methods) and 12 covariance estimators (unadjusted and 11 bias-adjusted estimators).
You can install the released version of geess from CRAN with:
install.packages("geess")And the development version from GitHub with:
# install.packages("devtools")
devtools::install_github("rtishii/geess")This is a basic example:
library(geess)
library(MASS)
# analysis of longitudinal count data usinBCg GEE method with Morel et al. covariance estimator
res <- geess(formula = y ~ trt + period + lbase + lage,
family = poisson, data = epil, id = subject,
repeated = period, corstr = "unstructured",
beta.method = "BCGEE", SE.method = "MB")
print(res)
#> Call:
#> geess(formula = y ~ trt + period + lbase + lage, family = poisson,
#> data = epil, id = subject, corstr = "unstructured", repeated = period,
#> beta.method = "BCGEE", SE.method = "MB")
#>
#> Model:
#> Family: poisson
#> Link: log
#> Correlation Structure: unstructured
#>
#> Estimation Method:
#> Regression Coefficients: BCGEE
#> Standard Errors: MB
#>
#> Number of observations: 236
#> Number of clusters: 59
#> Maximum cluster size: 4
#>
#> Coefficients:
#> (Intercept) trtprogabide period lbase lage
#> 1.8435 -0.0141 -0.0549 1.2397 0.6058
#>
#> Estimated Scale Parameter: 4.836
#> Number of Iterations: 7
#>
#> Working Correlation:
#> 1 2 3 4
#> 1 1.000 0.480 0.421 0.259
#> 2 0.480 1.000 0.624 0.331
#> 3 0.421 0.624 1.000 0.453
#> 4 0.259 0.331 0.453 1.000
#>
#> Convergence status: Converged
# hypothesis tests for regression coefficients
summary(res)
#> Call:
#> geess(formula = y ~ trt + period + lbase + lage, family = poisson,
#> data = epil, id = subject, corstr = "unstructured", repeated = period,
#> beta.method = "BCGEE", SE.method = "MB")
#>
#> Correlation Structure: unstructured
#> Estimation Method for Regression Coefficients: BCGEE
#> Estimation Method for Standard Errors: MB
#>
#> Coefficients:
#> Estimate Std.err Lower Limit Upper Limit Z Pr(>|Z|)
#> (Intercept) 1.8435 0.1531 1.5435 2.1435 12.0437 2.09e-33
#> trtprogabide -0.0141 0.1912 -0.3888 0.3607 -0.0736 9.41e-01
#> period -0.0549 0.0411 -0.1355 0.0257 -1.3345 1.82e-01
#> lbase 1.2397 0.1611 0.9238 1.5555 7.6927 1.44e-14
#> lage 0.6058 0.3157 -0.0129 1.2246 1.9191 5.50e-02
#>
#> Exp(coef) with 95% Confidence Intervals :
#> Exp(coef) Lower Limit Upper Limit
#> trtprogabide 0.986 0.678 1.43
#> period 0.947 0.873 1.03
#> lbase 3.454 2.519 4.74
#> lage 1.833 0.987 3.40
#>
#> Estimated Scale Parameter: 4.836
#> Number of Iterations: 7
#>
#> Working Correlation:
#> 1 2 3 4
#> 1 1.000 0.480 0.421 0.259
#> 2 0.480 1.000 0.624 0.331
#> 3 0.421 0.624 1.000 0.453
#> 4 0.259 0.331 0.453 1.000