FAiR 0.4-7 on 02-24-2011 to fix a potential clash with stats4
* Actually, the 0.4-5 and 0.4-6 fixed similarly minor issues
FAiR 0.4-4 on 01-12-2009 to fix another warning from R-devel
* Missed one warning last time
FAiR 0.4-3 on 01-09-2009 to fix warnings from R-devel
* Changes in the signature to the plot methods
* Fix manpages to comply with experimental parser
FAiR 0.4-2 on 10-06-2008 to fix documentation
* Only cosmetic changes
FAiR 0.4-1 on 10-03-2008 to fix breakage
* The acov slot of a manifest is temporarily of class "dMatrix"
FAiR 0.4-0 on 07-09-2008 in preparation for the POLMETH conference
* New virtual class, "manifest", that encapsulates the left-hand side of the model
- Inherited classes:
+ manifest.basic (superclass) if a covariance matrix is passed
+ manifest.basic.userW if a weight matrix is also passed
+ manifest.basic.data is the raw numeric data are available
+ manifest.data.ordinal if some variables are ordinal
+ manifest.data.ranks if the variables have be converted to ranks
+ manifest.basic.mcd if the MCD estimator was used
- Some methods are written for these classes, mostly just for manifest.basic
* New essential function, make_manifest(), to construct a "manifest.*" object
- Change in workflow: make_manifest(), make_restrictions(), Factanal() ...
- Some of the old functionality of make_restrictions() is now in make_manifest()
- New ways to estimate a covariance or correlation matrix
+ with missing data, using library(mlest)
+ with shrinkage, using library(corpcor) and FAiR_shrink() in Utils.R
+ Minimum Covariance Determinant (MCD), much improved over 0.2-0
+ based on Spearman rank correlations
+ polychoric and polyserial based on hetcor() in library(polycor)
+ bootstrapping the covariance matrix is available in many of cases
* New virtual class, "parameter", that represents a parameter to be estimated
- Inherited classes:
+ parameter.cormat for factor intercorrelation matrices
+ parameter.coef.{nl,SEFA,SEFA.nl} for primary pattern matrices
+ parameter.scale for scale matrices
- User internally with the new make_parameter() methods
* make_restrictions() is now S4 generic with lots of methods for parameter combinations
* New ways of specifying restrictions in make_restrictions()
- Much consolidation of pop-up menus making it easier to avoid some of them
- Easier to specify things at the command prompt via the new methodArgs argument
* New discrepancy functions, including and related to the Asymptotic Distribution Free (ADF)
- Elliptical, special case of ADF with common excess kurtosis
- Heterogenous Kurtosis (HK), special case of ADF
- Shrunk Heterogenous Kurtosis (SHK), special case of ADF
* New (correct) implementation of the embedded correlation factor analysis model
* New mapping rules in semi-exploratory factor analysis (SEFA)
- Put (some) zeros on high communality variables
- Either two or no zeros for each manifest variable
* New exact restrictions
- Constrain two or more coefficients to be equal to each other
- Arbitrary restrictions where one or more coefficients is some function of the others
* New inequality restrictions
- Order restrictions on factor contributions by row and / or column
- Designate the "best" indicators for any factor
- Prohibit some coefficients from being zero in SEFA
- Require a minimum binary distance between columns of the primary pattern matrix
* New inherited classes for EFA: restrictions.1storder.EFA and FA.EFA
* New (oblique) analytic transformation criteria from the GPArotation library
- Thanks to Coen Bernaards, Paul Gilbert, and Robert Jennrich for relicensing GPArotation
- Can use the reference structure or factor contribution matrix during transformation
- New convenience function, GPA2FA(), to convert a result from GPFoblq() or GPForth()
* New S4 generic functions and overhaul of previous S4 generic functions to ease development
- restrictions2model(), which is new and at the core of most everything
- restrictions2draws() and FA2draws(), which are new and used to estimate uncertainty
- restrictions2RAM(), which is new and converts to the RAM format of library(sem)
- cormat(), which is new and extracts correlation matrices
- uniquenesses(), which is new and extracts unique variances
- loadings(), which masks that in the stats library and extracts loadings
- gr_fitS4(), which is much improved but is now only used for MLE estimation
* Finally saw the light on S4 and converted all the S3 methods to S4 methods
* New export functions
- restrictions2Mathomatic(), which converts model to form used by Mathomatic
- write.cefa(), which converts preliminary loadings to form used by CEFA
* New tools for model comparison
- New functions, model_comparison() and paired_comparison()
- Incorporate contributions from Anne Boomsma and Walter Herzog
+ A correction to the test statistic suggested in Swain's dissertation
+ Improved RMSEA calculation
+ Steiger's gamma
+ McDonald's Centrality Index
- Many ADF-related test statistics suggested by Bentler and his coauthors
* plot() on an object that inherits from class FA now produces a Directed Acyclic Graph
* pairs() now depicts the correlation among the reference factors in the lower triangle
FAiR 0.2-0 on 01-31-2008 in honor of the date in the preface of Allen Yates' book
* Initial release for CRAN