Version 2.1
-------------------------------------------------------------------

The constrained optimization procedure is now done in two stages for LEXPIT for the linear and expit components. This has the benefit of reducing constraints to linear piece where all inequalities will be linear in the parameters.

Naming of ci changed to confint.

The hosmerlem function has been changed to gof.

The gof.pearson function has been added to compute Pearson's goodness-of-fit test.

A revised vignette reflects the changes in 2.1.


Version 2.2
-------------------------------------------------------------------

Added warn argument to suppress warnings thrown when testing for feasible parameters.

Function excess.risk added. Computes binned excess risk by a grouping variable. Useful for diagnosing functional relationship between risk and explanatory variable.

Correction to variance calculation to include case status as a stratification variable for case-control studies.

Revised vignette includes description of use of the excess.risk function.

Added likelihood ratio test for hypothesis testing with cohort data (method LRT).

Added leverage and displacement functions for blm and lexpit models to diagnose the model fit.

Version 2.3
-------------------------------------------------------------------

Fixes to vcov method to handle big matrices.

Correction to gof function and addition of adjustment for weighted estimates.

Version 2.4
-------------------------------------------------------------------

Added initialization parameters in influence methods.

Modified summary method to shorten print out of t and p-values.

Added aarp data set.

Version 2.4.1
-------------------------------------------------------------------

Removed dev directory

Version 2.4.2
-------------------------------------------------------------------

Added risk.exposure.plot; removed excess.risk function; updated vignette with example for this plot

Version 2.4.3
-------------------------------------------------------------------

Corrected bug in stratification variable in variance calculation for blm (function: vcov.influence.blm.strata); update predict method so response variable is not necessary

Version 2.4.4
-------------------------------------------------------------------

Added formula accessor; Made output of predict a vector; Use printCoefmat in the summary function; Created crude.risk function; Added log-likelihood method; Added dependency on stats4; References and CITATION to JSS paper