-
Notifications
You must be signed in to change notification settings - Fork 13
GEE and QIF for clustered data regression
Package geepack is one of the major tools in R for generalized estimating equations (GEE). The C++ code in the geepack package depends on a set of headers called TNT. So many years have passed and much more convenient tools such as Rcpp and RcppArmadillo are available, which would make the code much easier to maintain and extend. Further, it is also time to add new functions to incorporate the recent advances such as quadratic inference functions (QIF), user-specified link/variance functions, working correlation structure selection, regularized estimation, among others.
Package gee is based on Vincent Carey's legacy C code. Package geeM is a pure R implementation relying on the Matrix package for sparse matrix operations. Package multgee does GEE for nominal or ordinal responses.
- A new implementation of the basic GEE using RcppArmadillo.
- Allows user-specified link function and variance functions defined in R.
- An implementation of QIF using RcppArmadillo.
- Regularized estimation for GEE and QIF.
- Tests and comparison with existing implementations.
- One or more vignettes to document them.
GEE and QIF are indispensable tools for marginal regression modeling of clustered data with wide applications in many disciplines where clustered data are needed to be analyzed.
Students, please contact mentors below after completing at least one of the tests below.
- Jun Yan [email protected] is an author of R packages geepack, copula, and some more.
- Yixuan Qiu [email protected] attended GSOC twice. He is the author of R packages showtext, RSpectra, recosystem, prettydoc, among others.
Students, please do one or more of the following tests before contacting the mentors above.
- Easy: .
- Medium: Write a C++ function that can be called from within R to do a generalized linear regression.
- Hard: Write a C/C++ function, which accesses a family object (e.g., binomial, gaussian, etc.) in R and demonstrates its link function, variance function, and mu.eta function.
Students, please post a link to your test results here.
https://github.com/AnanyaTyagi/gsoc2018.git
- Name: Yuze Zhou
- Solutions: https://github.com/zhouyuze/gsoc_test