bife news

Changes in version 0.7.3

• Replaced 'qr()' by 'lm.fit' to fix memory issue

• Updated maintainer email address and removed CPP11 restriction (#13 @zeileis)

Changes in version 0.7.2

• Replaced NEWS.txt with NEWS.Rd.

• Improved the finite sample performance of bias corrections for the average partial effects in case of perfectly classified observations.

• The default option 'n_pop' in get_APEs() has been changed. Now the estimated covariance consists of the delta method part only, i.e. correction factor = 0.

Changes in version 0.7.1

• Step-halving is now similar to glm.fit2().

• Fixed an error in the covariance (influence function) of get_APEs().

• Added reference for the bias correction in 'citation()'.

• Updated some references in the documentation and vignette.

• Fixed some typos in the documentation and vignette.

Changes in version 0.7

• Added a generic function logLik() that returns the sum of the log-likelihood.

• 'X_new' in predict() now also permits data.frames.

• bife() now permits to expand functions with poly() and bs().

• Added option 'sampling_fe' to get_APEs() to impose simplifying assumptions when estimating the covariance matrix.

• Fixed an error in get_APEs() related to the estimation of the covariance.

• Fixed a bug in the internal function used to estimate spectral densities.

Changes in version 0.6

• Major source code cleaning and revision.

• Vignette renamed and revised.

• The asymptotic bias-correction is ported to a post-estimation routine bias_corr().

• Replaced the bias-correction of Hahn and Newey (2004) with the one of Fernandez-Val (2009), i.e. refined bias expressions plus dynamic feedback possible.

• The post-estimation routine apeff_bife() was renamed to get_APEs() and also returns an estimate of the covariance using the delta method.

• Fixed a bug where the number of perfectly classified observations was not reported correctly. This bug also caused the estimates of the average partial effects to be unreliable for small time horizons.

• Some performance tweaks.

• Removed 'acs' data to reduce overall package size.

Changes in version 0.5

• Changed the handling of factor() in the formula interface. This should now be more in line with what users expect.

• fixed() is removed since it is redundant now.

• Realigned print.summary.bife().

Changes in version 0.4

• Fixed a memory-access error in 'bife.cpp'.

Changes in version 0.3

• Added a generic function predict().

• Additional downgrades to ensure compability with g++ 4.4.x.

• Downgraded compiled code from C++11 to C++98. This ensures compability with older compilers.

• Removed jackknife bias-correction from bife() and apeff_bife().

• apeff_bife() now uses the full sample instead of a sub-sample of indiviuals with a varying response.

• Changed default argument 'discrete' of apeff_bife() to 'NULL'.

• Changed default argument 'bias_corr' of apeff_bife() to '"ana"'.

• Fixed an issue where vcov() was not able to distinguish between corrected and uncorrected coefficients.

Changes in version 0.2

• Added a function fixed() to model additional fixed-effects. See documentation for further details.

• Re-structured compiled code to optimize package size.

• Changed some function and argument names to common naming convention. For instance apeff.bife(..., bias.corr = "ana") is now apeff_bife(..., bias_corr = "ana").

• Downgraded compiled code from C++11 to C++98. This ensures compability with older compilers.

• Added comments and renamed variables such that they are in line with the notation of the corresponding working paper.

• Fixed some typos in documentation.

• Fixed an issue where bife() was not able to fit a model with just one explanatory variable.

• Corrected computation of degrees of freedom.

Changes in version 0.1

• Initial release on CRAN.