Flic models7/4/2023 (2010): "Bias-reduced and separation-proof conditional logistic regression with small or sparse data sets", Statistics in Medicine 29:770-777 doi:10.1002/sim.3794. (2008): "Avoiding infinite estimates of time-dependent effects in small-sample survival studies", Statistics in Medicine 27:6455-6469 doi:10.1002/sim.3418. (2006): "A comparative investigation of methods for logistic regression with separated or nearly separated data", Statistics in Medicine 25:4216-4226 doi:10.1002/sim.2687. (2003): "Fixing the nonconvergence bug in logistic regression with SPLUS and SAS", Computer Methods and Programs in Biomedicine 71:181-187 doi:10.1016/S0169-2607(02)00088-3. (2002): "SAS and SPLUS programs to perform Cox regression without convergence problems", Computer Methods and Programs in Biomedicine 67:217-223. (2002): "A Solution to the Problem of Separation in logistic regression", Statistics in Medicine 21:2409-2419 doi:10.1002/sim.1047. (2001): "A Solution to the Problem of Monotone Likelihood in Cox Regression", Biometrics 57(1):114-119 doi:10.1111/j. This work was supported by the Austrian Science Fund (FWF), award I-2276 and by the European Commission's MSCA programme, grant agreement number 795292. It differs from the method described in the paper only by the missing small-sample correction of the empirical (sandwich) variance estimation. The macro builds on PROC LOGISTIC and PROC GEE. This macro implements the single-step augmented GEE (augGEE1) approach for fitting GEE with a binary outcome described by Geroldinger, Blagus, Ogden and Heinze (2022). The macro implements the method proposed in Joshi et al (2021). Multiple, equally-structured data sets can be processed with very efficient use of BY-processing. The macro builds on iterated calls of PROC GENMOD. With this macro, the Firth and FLIC/FLAC methods can be used with Poisson and Negative Binomial regression. The accompanying technical report contains instructions on how to implement analyses for 1:m matched case-control studies with Firth-correction using the macro. It is based on FORTRAN code and an external routine (either included as EXE or DLL) which must be made invokable from the SAS macro. Implements Firth's correction for Cox regression as described by Heinze and Schemper (2001). It uses SAS/PROC LOGISTIC to compute the conditional distribution of sufficient statistics which can computationally burdensome. Implements the conditional Firth-corrected logistic regression methods described in Heinze and Puhr (2010). Recently, van Calster, van Smeden, de Cock and Steyerberg (2020) showed that the FLIC method can yield calibration slopes which have mean squared error smaller than competing methods that use cross-validation to tune penalty parameters such as the Lasso or ridge regression. With rare events, Firth correction can lead to inflated average predicted proabilities such that predictions are biased high. Unlike the default Firth correction, with FLIC and FLAC it is guaranteed that the average predicted probability is equal to the observed event rate. These methods are particularly interesting for predicting with penalized logistic regression. This macro implements the FLIC and FLAC methods as described by Puhr, Heinze, Nold, Lusa and Geroldinger (2017). It also computes profile penalized likelihood confidence intervals as described by Heinze and Schemper (2002), Heinze and Ploner (2003), Heinze (2006), and Mansournia, Geroldinger, Greenland and Heinze (2018). Unlike the implementation of Firth's correction in SAS/PROC LOGISTIC, this macro is also able to provide p-values based on penalized likelihood ratio tests for each regression coefficient. The 'old' SAS macro to fit logistic regression models using SAS/PROC IML code. Special macros are available to implement the FLIC and FLAC methods of Puhr et al (2017) doi:10.1002/sim.7273. Here we provide our SAS-macros to fit Firth-corrected regression models, in particular logistic, conditional logistic and Poisson regression models. SAS-macros for Firth's corrected logistic, conditional logistic and Poisson regression, FLIC and FLAC methods Description
0 Comments
Leave a Reply. |