NEWS | R Documentation |

New argument 'cores' => added multi cores with openMP. Particularly efficient for the Negative Binomial and the Logit models.

Added a fixed-point acceleration algorithm for 3+ clusters and also to obtain the derivatives.

Cluster coefficient precision is dynnamically adjusted for complex problems.

Faster matrix creation.

[Design] Argument noWarning is dropped.

[Design] clearer message when clusters are removed.

[Bug] corrected undue information reporting.

[Bug] corrected important bug that made some estimations fail.

[Help] Added more accurate information on the returned objects.

[c++ code] Fixed overloading problem.

Improved help pages.

Code more robust to overfitting situations (in particular for the Logit case).

Added an acceleration algorithm to get the fixed-point for 2 clusters.

New collinearity warnings.

Algorithm is greatly improved.

Two new likelihood models: Negative Binomial and Gaussian.

Any number of clusters is supported.

The functions res2table/res2tex to export the results are greatly enhanced.

cluster-robust covariance: the method to compute it has been improved (more robust to rounding problems).

Computing standard errors: up to four-way clustering is supported.

The function designing has changed.

There is only one function called femlm that performs "normal" estimations and "non.linear" in parameters estimations when adding the argument NL.fml.

clusters can now directly be inserted in the formula without using the arument "cluster"

Name change: the argument "dummy" becomes "cluster".

Bug related to the handling of the upper and lower bounds is now corrected.