2014-05-26
* numDeriv now correctly passed in case some VP are refitted,
when fitting failed with analytical gradient.
2014-05-14
* added numDeriv argument to fit functions.
2014-05-08
* added determaint_c.cpp which was missing. A clear bug...
2014-01-18
* removed bug that fit.model stopped too often
when check.model was TRUE.
2014-01-18
* removed bug in get.mpt.fia
2014-01-03
* removed bug in lbmpt.to.mpt (captial letters)
* added restrictions argument to make.mpt.cf
2014-01-02
* added function lbmpt.to.mpt
2013-12-14
* fit.mpt and fit.model now work with args.fia
2013-12-13
* added mConst argument to bmpt.fia for avoiding
numerical underflows. See:
http://stackoverflow.com/q/20529407/289572
* added mConst or args.fia to rel functions.
* package Brobdingnag now Imported for same issue.
2013-11-24
* added check.model to gen.predictions et al.
2013-11-06
* removed bug when only fitting a model with only one
free parameter to multiple data sets (reported by
Alberto Mirisola)
* if FIA fails fit.mpt gives a warning now (instead
of failing)
* if check.model = FALSE (fit.model) the check of
whether or not the parameter are within bounds is
now also suppressed.
2013-11-03
* added check.model argument to fit.model.
2013-10-28
* check.mpt shouldn't fail if predictions are NA.
2013-10-28
* bmpt.fia now uses the absolute of the determinant
(as did the original code by Wu et al.).
* removed restriction of constant to within 0 and 1.
2013-06-26
* added reasonable message when objective function failed.
* changed warnings to messages.
* now everything should work with Hessian (hopefully)
2013-06-25
* last changes broke code, should be better now
2013-06-24
* removed bug that Hessian function was invoked
even though use.hessian = FALSE.
2013-05-24
* needs R (>= 2.14) same as RcppEigen
* minor correction to bmpt.fia.Rd
2013-05-23
* sfLibrary now called with only.character = TRUE
(removes R CMD check NOTE).
2013-05-22
* fixed line lengths in .Rd files
2013-05-17
* updated documentation of fit.mpt (added
warning for possibly inconsistent FIA penalties)
2013-05-11
* removed bug in make.mdt with data.frames
2013-05-09
* added n.smaller.05 column to select.mpt
* added note about factorial constant (fit.mpt, fit.model)
2013-05-07
* multicore fitting now uses load balancing (fit.mptinr)
2013-04-01
* FIA calculation can now use multicore (in bmpt.fia)
* multicore FIA also available via get.mpt.fia, fit.mpt,
and fit.model.
2013-02-24
* FIA calculation uses RcppEigen (thanks to
Christian Mueller)
2013-02-15
* added feature to specify lower and upper bound
for starting values (must be two element list,
first is lower, second is upper).
2013-02-10
* modifed documentation for FIA (extra brackets warning).
2013-02-02
* aggrgeated G^2 only if output == "full"
2013-01-27
* added CITATION file
* added aggregated G^2 when output = "full" (select.mpt)
2012-09-14
* resolved bug when using inequality and crossed equality
restrictions (see bugs/resrictionsbug.order.R)
2012-09-03
* select.mpt now shows g.square, df, and p
* select.mpt has argument dataset
2012-08-30
* added function plot.prediction.
* added FIA.penalty to output with FIA.
* removed bug when select.mpt from models with fit
fit.aggregated = FALSE
2012-07-24
* make.mdt now uses dimnames(data)[[1]] as index (with warning).
2012-07-20
* make.mdt can now handle a matrix or data.frame, updated .Rd file
* reactivated examples for the gen.data examples (gen.data.Rd)
* make.mdt now makes mdt files that HMMTree can handle
2012-07-18
* disabled examples for the gen.data examples (gen.data.Rd)
2012-07-16
* added sample.data function for nonparametric bootstrap
* added data argument to gen.data. Now, either data or
n.per.item.type need to be non-null (but not both).
2012-07-10
* now comments in model and restrictions files are treated like
in normal R (i.e., only everything after # is ignored).
Changed: .read.MPT.model() and .read.MPT.restrictions()
2012-07-09
* added functions gen.predictions and gen.data
* MPTinR is now byte compiled.
2012-05-08
* resolved bug when obtaining FIA for individual data sets
2012-05-02
* resolved bug when obtaining FIA from model specified via textConnection.
2012-04-26
* fit.mptinr now checks if there is a gradient/hessian function and
sets use.gradient/use.hessian FALSE otherwise.
2012-04-18
* copy editing on documentation.
* select.mpt(): renamed n.model.df to n.independent.categories
2012-04-16
* more work on bug when paramaters in restrictions are not in model
(work from yesterday introduced more problems)
added function .check.restrictions() which also reads the restrictions
* added an aggregated example to fit.mpt.Rd
2012-04-14
* data element now contains indidivudal and aggregated data
* more work on bug when paramaters in restrictions are not in model
2012-04-13
* removed warning when cell counts are < 5
* restrictions.filename can now be a list of characters
* worked on examples (textConnections moved in the call to fit.mpt)
2012-03-28
* fixed bug when restricted parameters were not present in the model
2012-03-19
* model.info does now return the rank of the Fisher Matrix.
* summary.llks was renamed to fitting.runs
2012-02-25
* fitting functions(fit.mptinr, fit.mpt, fit.model) contain the option to not
fit the aggregated data for multiple datasets.
2012-02-22
* check.mpt now also displays number of categories and number of df for the model.
* MPTinR-package.Rd updated to current version number and more functions.
2012-02-07
* select.mpt now rounds information criteria before determining number of
best fits (also added this to select.mpt.Rd)
2012-02-06
* resolved small bug for select.mpt (wouldn't show aggregated FIA results
when output = "full")
* added show.messages parameter to fit.model, fit.mptinr and fit.mpt
(including their documentation)
* refitting for failed nlminb now uses multicore.
2012-02-04
* small changes in Rd files for fit.model and fit.mptinr
2012-01-31
* resolved small bug when fitting using mutlicore = "individual"
* changed Description
* changed warning messages when optimization did not converge successfully.
2012-01-30
* resolved small bug in calculating the FIA (moved routine to calculate
the FIA to the wrapper functions fit.mpt and fit.model, changed argument
fia to fia.df in fit.mptinr). Changed fit.mptinr.Rd accordingly.
* reading models should now work with connections (changed .get.mpt.model so that
connections are always read as "easy").
* added function make.dt with help page (in make.eqn,.Rd)
* added examples in fit.mptinr and fit.model
* changed default starting values to c(0.1, 0.9).
2012-01-27
* first release of MPTinR 0.8.x. changed fit.mpt and fit.model to be wrappers
around fit.mptinr, which now uses nlminb.
* fit.mpt.old is the old version of fit.mpt using otpim.
* changed the rd files accordingly to reflect the changes above.
2012-01-18
* fixed but in fit.model()'s lower and upper.bound arguments. Now works
2012-01-12
* changed .find.mpt.params to rely on all.vars()
* adapted fit.mpt() (creating parameters from inequality restrictions) to
deal with the new .find.mpt.params() function (i.e., inequalities are
parse()d prior to calling .find.mpt.params).
* added fit.model(), a copy of fit.mpt() with parameter bounds as an argument
2011-06-30
* R/fit.mpt.R: now starting value parameter works
* Small bug removed when fitting one row from a data.frame (now, does not set multiFit to TRUE)
2011-06-29
* R/fit.mpt.R: fixed crashing when hessian of aggregated data was ill defined
2011-06-28
* R/fit.mpt.R: added starting.values parameter.
* R/fit.mpt.Rd: added description for the starting.values parameter.
* R/fit.mpt.R: internal changes: replaced length.tree with n.params.
2011-06-14
* R/bmpt.fia.R: if theta is a scalar, it now is a matrix.