2015-06-12 Giulio Bottazzi
* ALL: Tagged with CVS using
#cvs tag gbutils-5-6-6
* configure.ac: Source package version updated to 5.6.6.
* gbstat.c: added the computation of the mode using the Half
Sample Method by Robertson and Cryer.
2015-06-06 Giulio Bottazzi
* ALL: Tagged with CVS using
#cvs tag gbutils-5-6-5
* gbkreg2d.c, gbkreg.c, gbker2d.c, gbhisto2d.c, gbacorr.c,
gbstat.c : fixed typos
2015-05-30 Giulio Bottazzi
* gbget.c: set the expression 'x0l1' to the row number of the
previous row, consistently with general notation.
* configure.ac: added automatic manual creation for gbget
2015-05-29 Giulio Bottazzi
* tools.c: updated copyright dates
* configure.ac: Source package version updated to 5.6.5.
* Makefile.am: removed redundant script gbkeys2values: it gets
replaced by gbconvtable.
* gbconvtable, gbenv, Makefile.am: added man pages for gbconvtable
and gbenv.
2015-05-12 Giulio Bottazzi
* ALL: Tagged with CVS using
#cvs tag gbutils-5-6-4
* gbhill.c: updated and improved the comment in the source code
* gbget.c: added information about the size of the final matrix in
verbose mode.
2015-05-06 Giulio Bottazzi
* configure.ac: Source package version updated to 5.6.4.
* Makefile.am: removed 'debian' directory from the distributed
package. This is required for the inclusion in the official debian
repositories.
2015-04-22 Giulio Bottazzi
* gbhill.c: improved help and error messages
2015-03-29 Giulio Bottazzi
* configure.ac: removed any reference to 'flex' apart the initial
"AM_PROG_LEX" which seems to be useful on Cygwin (but not on linux
systems). Flex should be in general checked by the libmatheval
package. Corrected dependencies of the gbget program. Source
package version updated to 5.6.3. Tagged with CVS using:
#cvs tag gbutils-5-6-3
* Makefile.am: correct dependency for gbget; the explicit
specification of flex library in the '_LDADD' declarations is now
activated only for Cygwin.
* gbkreg.c : corrected typo in help message
2015-03-25 Giulio Bottazzi
* gbmave.c: added default value for option '-s' to help message.
2015-03-05 Giulio Bottazzi
* gblreg.c (main): fixed bug in option '-O 2': the estimated
coefficient were wrongly printed on the first line.
2015-01-08 Giulio Bottazzi
* ALL: present version becomes 5-6-2 (tagged with CVS) using
#cvs tag gbutils-5-6-2
This can be recovered using
#cvs checkout -r gbutils-5-6-2
2015-01-07 Giulio Bottazzi
* configure.ac, debian/changelog: version updated to 5.6.2; used
'dch -v 5.6.2-1' to update debian/changelog
* Makefile.am: updated to reflect new documents structure
* gbzscore, gbdenan, gbrobreg, gbrobareg: obsolete utilities
removed from the package
* gbdummyfy, Makefile.am: added man page
* gbrand.c, gbker.c, gbinterp.c, gbhisto2d.c, gbboot.c, gbglreg.c:
error messages are now correctly printed to stderr
2015-01-06 Giulio Bottazzi
* gbinterp.c, gbrand.c: warning messages are now correctly printed to stderr
2014-12-27 Giulio Bottazzi
* README, debian/docs, doc: added directory containing
documentation files cygwin_install.pdf, cygwin_install.txt,
gbget.pdf, gbget.txt, intro.pdf, intro.txt, overview.pdf,
overview.txt
2014-09-26 Giulio Bottazzi
* README: updated list of source files
2014-09-25 Giulio Bottazzi
* tools.c: added 'finalize_program' to free all variables
inizialized in 'inizialized_program'
* tools.h: exported variable 'NL'
* tools.c, tools.h, gbquant.c, gbget.c, gbtest.c: added the
function "denan_pairs"; the function "data_denan" has been renamed
"denan_data" for consistency
2014-09-23 Giulio Bottazzi
* gbtest.c (run_twosamples): changed the output of two sample
tests in the case of many columns. Now point estimates are
returned in the lower triangle irrespective of option -p.
2014-09-10 Giulio Bottazzi
* Makefile.am: the list of files to be included in the package
has been updated
2014-09-03 Giulio Bottazzi
* ALL: present version becomes 5-6-1 (tagged with CVS) using
#cvs tag gbutils-5-6-1
This can be recovered using
#cvs checkout -r gbutils-5-6-1
* debian: updated file in the debian dir
* gbhisto.c: added option '-S' to tune the midpoint calculation;
added two cases to otpion -M to print cumulated distribution;
fixed the output of some warning messages
2014-09-02 Giulio Bottazzi
* gbhisto.c: added the description of the bin midpoint in the help
message
* gbacorr.c: added to the CVS repository
2014-08-27 Giulio Bottazzi
* gbacorr.c: added optiopn -p to compute confidence intervals
using Fisher transform
* configure.ac, Makefile.am: added 'gbacorrr' among the utilities
depending on gsl for installation. Fixed mis-allignemt of the
rules to create man pages.
2014-08-26 Giulio Bottazzi
* tools.h: removed redundant initializaton of optarg ,optind,
opterr and optopt;
* Makefile.am: added explicit linking option for gbget
2014-08-20 Giulio Bottazzi
* gbplot: added an example using the 'set table' option; removed
uneccessary '\' in the help message
2014-06-25 Giulio Bottazzi
* ALL: present version becomes 5-6 (tagged with CVS) using
#cvs tag gbutils-5-6
This can be recovered using
#cvs checkout -r gbutils-5-6
* debian/control (Suggest): added gnuplot to suggested packages
2014-06-19 Giulio Bottazzi
* gbnlprobit.c (main): added a one line header to verbose messages
2014-06-18 Giulio Bottazzi
* gbnlprobit.c (main): fixed argument of 'sizeof' in my_alloc
statements
2014-06-17 Giulio Bottazzi
* tools.c (readblock): fixed nasty bug in the procedure with wich
missing columns are filled with NAN
2014-06-12 Giulio Bottazzi
* ALL: version updated to 5.6
* gbplot: added man page
2014-06-07 Giulio Bottazzi
* gbmave.c: compute moving average for each column of data;
previously the data from all columns were pooled
2014-06-06 Giulio Bottazzi
* README: added mention of gbacorr and gbxcorr
* ALL: implemented the long option '--version' and '--help' for
all the executable, but not yet for the scripts. Now the help
messages are sent to sdtout, in compliance with GNU coding
standards.
* tools.c, tools.h: added structure for getopt_long
2014-06-05 Giulio Bottazzi
* Makefile.am, configure.ac : added help2man support. The man
pages are added to the list as the revision of the help messages
progresses.
* gbker.c, gbget.c (main): Updated the -h message to comply with
help2man syntax
2014-06-03 Giulio Bottazzi
* gbxcorr.c: updated gbxcorr to conmpute cross-covariance and
cross-correlation coefficients with and without the removal of the
mean
2014-05-18 Giulio Bottazzi
* gbacorr.c : added program 'gbacorr' to compute the
autocorrelogram (or the cross-autocorrelogram) of a series of
observations, column-wise.
2014-05-16 Giulio Bottazzi
* debian: added the 'debian' directory that contains all
information necessary to create the package. After 'autoreconf',
'./configure' use 'make dist' to obtain a source tarball which can
be used to create the deb packages. Remember to change the name of
the source package from gbutils-version.tar.gz to
gbutils_version.orig.tar.gz. Extract the archive, move inside the
source tree and issue 'debuild -us -uc -b'
* Makefile.am (EXTRA_DIST): directory 'debian' added to list of
distributed files
2014-05-15 Giulio Bottazzi
* ALL: version set to 5.5.1
* gbker.c (main): added option -w to select the boundaries of the
gird. Modified method -M 2 for a more precise boundaries
management. Using '-M 2 -w x,x -n 1' it is possible to compute the
density in the point x.
* configure.ac: updated syntax
* tools.c (gbutils_header): updated header output
* gbnlpolyit.c: improved help message
2014-03-31 Giulio Bottazzi
* gbtest.c (WMW_pscore): changed the text of WMW verbose output
2014-03-20 Giulio Bottazzi
* gbdist.c: added short description of the procedure
2014-02-26 Giulio Bottazzi
* gbnlreg.c (main): corrected bug that printed two times the
number of observations
2014-02-20 Giulio Bottazzi
* ALL: version set to 5.5 (consolidation release)
* gbnlreg.c: added R^2, F statistics and one-sided significance
for the OLS model (-m 0). The baseline model is NOT the model with
the coefficients (or parameters) equal to zero, but the model with
the coefficients equal to the values specified in the initial
condition.
* gbnlpanel.c, gbnlmult.c, exponential_gbhill.c, paretoI_gbhill.c,
gbhill.c, gbker.c, gbmstat.c, gbgcorr.c, gbboot.c: checked with
-Wall; removed unused variables.
* gbnlprobit.c: found potential bug in score minimization
* multimin.c: Added two new GSL alogithms for minimization without
derivatives. Updated version to 1.2
2014-01-17 Giulio Bottazzi
* gbnlreg.c: corrected bug in the LL of gaussian model
2013-12-20 Giulio Bottazzi
* gbnlreg.c : implemented the report of various information
criteria in the verbose mode. The statistics reported are : AIC,
AICc, BIC and HQCin the case of LS regression
2013-11-23 Giulio Bottazzi
* ALL: version set to 5.4.5 (bugfix release)
2013-11-22 Giulio Bottazzi
* gbdist.c: avoid to print multiple values for consecutive
identical observations. Only the last one is printed. Added the
option '-i' (for identical) to recover previous behaviour. in the
function printdist the type of the counter has changed form int to
size_t
2013-10-23 Giulio Bottazzi
* gbnlmult.c, gbrand.c, gbnlqreg.c : fixed minor bugs
* tools.c, tools.h, gbkreg2d.c, gbker2d.c, gbker.c, gbkreg.c,
gbmodes.c, gbnear.c: the inline specification has been removed
from the definition of the different types of kernel and the
functions have been moved in tools.c/tools.h
2013-02-01 Giulio Bottazzi
* ALL: version set to 5.4.4 (bugfix release)
2013-01-21 Giulio Bottazzi
* tools.c (readblock): added separate warnings for underflow and
overflow.
2012-11-10 Giulio Bottazzi
* gbdummyfy: corrected bug in the selection of the column to
delete.
2012-11-07 Giulio Bottazzi
* tools.c, gbenv.c: added two new environment variables,
GB_OUT_SEP and GB_OUT_NL to set the output field and record
separators respectively.
2012-11-06 Giulio Bottazzi
* gbnlreg.c: removed the condition |dF|>0 in the convergence
condition of the OLS fitting. Indeed it could be |dF| == 0 which
is perfectly fine! Notice that with linear functions, the second
iteration of the minimizer fail automatically (no way of improving
upon the first iteration!) but this does not generate any trouble
in gbnlreg, apart the printing of a somehow cryptic message
"cannot reach the specified tolerance in F" when verbosity is set
to high.
* configure.ac: updated the check for the gsl, according to the
recommendation in the gsl manual
2012-10-29 Giulio Bottazzi
* tools.h, tools.c: replaced the occurrences of "GBFILE *" with
GBFILEP, which is already a pointer. It is substituted by 'gzFile'
if the zlib are found on the system or by "FILE *" if they are
not. This is actually the fixing of a bug, that went unnoticed in
the old zlib versions but is now rightly spotted by the compiler.
2012-10-25 Giulio Bottazzi
* tools.c (readblock): added different warnings for different
errors when converting string to double.
2012-10-24 Giulio Bottazzi
* gbnlreg.c: added two relevant statistics to the LAD estimation
(-M 1), namely the D statistics for inference and the coefficient
of determination R2. The asymptotic probability to observe a D
statistics as large as observed under the null (that is with
parameters equal to their initial conditions) is also
reported. These quantities are reported if the option -v is
greater than 1.
2012-10-02 Giulio Bottazzi
* gbnlreg.c: fixed bug in the conditions implying var-covar matrix
computation
2012-10-01 Giulio Bottazzi
* gbnlreg.c (main): computation of var-covar matrix only when
needed; print warning message if the maximum number of iterations
is reached without optimization success; implemented a restart
algorithm for simplex minimization, i.e. -M 1 and -M 2; new
command line options -I and -E.
2012-09-25 Giulio Bottazzi
* gbconvtable: fixed a HUGE bug: replacement should be performed
also if the column considered is not the last one
2012-07-09 Giulio Bottazzi
* gbnlpolyit.c (main): Cox R^2 correctly renamed Effron R^2.
2012-07-04 Giulio Bottazzi
* ALL: version set to 5.4.3 (bugfix release)
* gbconvtable, gbdummyfy, gbzscore, gbdenan, gbplot: improved
script portability by substituting '=' for '==' and replacing
double square brackets test operator '[[' with single square
bracket '['. The option '-E' has been removed from 'echo'
instances and the help message is printed using an here-doc syntax
instead than an 'echo' command. The 'if-then' syntax has been
replaced with a 'case' construct for the parsing of the command
line.
2012-06-23 Giulio Bottazzi
* gbinterp.c: in points outside the input range the interpolation
of the curve is set equal to the nearest value while the
derivatives are set to zero. A warning is printed.
2012-05-28 Giulio Bottazzi
* ALL: version set to 5.4.2 (bugfix release)
* configure.ac: added version number to the final greeting;
updated checking for (f)lex
2012-05-15 Giulio Bottazzi
* gbhill.c, gaussian_gbhill.c, paretoI_gbhill.c,
paretoIII_gbhill.c, exponential_gbhill.c: removed the
"unconditional" likelihood (previous options -M 1 and -M 4)
2012-05-14 Giulio Bottazzi
* gbtest.c (main): updated names of the statistics according to
Stehphens' 1974 paper on J. Am. Stat. Ass.
2012-03-23 Giulio Bottazzi
* gbrand.c (main) : implemented binary output for the univariate
continuous distributions
* gbtest.c (main) : implemented binary input
* gbhill.c (main) : implemented binary input/output; the binary
output is for residuals
* tools.c, tools.h: reorganized the list of functions and added
function 'printcol' and 'printcol_bin' which print a single column
of output
* gbfun.c (main), gbget.c (main), gbmstat.c (main) : fixed the
option -b for binary input/output
2012-02-27 Giulio Bottazzi
* tools.c (readblock): properly produce NAN when a conversion from
string to float fails; print a warning if some conversion failed.
2012-02-14 Giulio Bottazzi
* multimin.c: corrected the count of iterations. Now when the
maximum number of iterations is set to zero, the program correctly
does not perform any minimization and returns the initial values.
* gbhill.c: included the Renjy transformation based on subsequent
spacings. Still to be tested.
2011-12-21 Giulio Bottazzi
* ALL: version set to 5.4.1 (bugfix release)
* gbhill.c: improved the treatment of Renjy transform for the
goodness of fit, introducing "artificial" distribution with mass
only on the range covered by the sub sample used in the
estimation.
2011-10-24 Giulio Bottazzi
* gbxcorr.c (main): added utility to compute the cross-correlation
matrix.
2011-10-04 Giulio Bottazzi
* ALL: present version becomes officially 5.4 and it is released
* gbget.c, gbnlpolyit.c: removed unused variable
* gbtest.c: added explicit de-allocation of vectors
* gbhisto.c: corrected bug in rage calculation in the case of
many samples
* tools.c (my_alloc): changed any instance of alloc to my_alloc.
2011-10-03 Giulio Bottazzi
* gbget.c (new_matrix): removed unused variable.
* gbhisto2d.c (main): removed unused function
'printhist2d'. Variables 'xsteps' and 'ysteps' changed type from
int to size_t.
* gbtest.c (KW_test, FP_test, RHO_test): corrected bug in memory
allocation.
* tools.c (readblock): corrected bug in data loading: missing
values weren't properly set to NAN. Print the warning for missing
values only once.
2011-08-02 Giulio Bottazzi
* gbnlpolyit.c: corrected a bug in the bootstrap calculation of
standard deviation of marginal elasticities and marginal
effects. Added the printing of p-scores together with standard
deviations when the option -v is specified. Added three different
pseudo R^2 printed with the verbose option.
2011-08-01 Giulio Bottazzi
* gbhill.c: Improved option -O 4, now the uniform transformation
of Renjy variables is printed for any kind of estimator. Removed
option -O 5 because it is redundant. Now the number of parameters
is kept fixed and NAN are added where appropriate.
2011-07-21 Giulio Bottazzi
* gbglreg.c: added option -O 4 to print fraction of explained
variance associated to each explanatory variables.
2011-07-19 Giulio Bottazzi
* gbnlpolyit.c: corrected bug in the determination of the number
of occupancy classes. Added explicit calls to free().
2011-07-18 Giulio Bottazzi
* configure.ac: removed residual references to FFTW; introduced
check for lex/flex library necessary to have a working matheval
installation.
* NEWS: start to update the file. It was not updated after
revision 5.0
* README: removed references to FFTW
2011-07-08 Davide Pirino
* gbhill.c: Added a new warning in the case of an exponential or
pareto1 distribution for the lower tail estimation.
* exponential_gbhill.c: Added varcovar for the lower tail estimation.
* pareto1_gbhill.c: Added varcovar for the lower tail estimation.
* gaussian_gbhill.c: Added varcovar for the lower tail estimation.
2011-07-13 Giulio Bottazzi
* ALL: present version becomes 5.4; but some bug fixes in gbhill
are necessary
* gbkreg.c, gbker.c, configure.ac : dropped dependence on FFTW
library
2011-07-09 Giulio Bottazzi
* multimin.c: added transformation number 9 for semi-open interval
(-infty,xmax) to be used, for instance, in gbhill.
* gbhill.c: moved var-covar matrices in the distribution-specific
files; changed default options of the minimization routines to use
BFGS-2; changed the transformation used for the exponential
distribution.
2011-07-08 Davide Pirino
* exponential_gbhill.c: bugs fixed.
2011-07-05 Giulio Bottazzi
* gbplot: fixed a bug in option '-p' and added verbose output with
option '-v'.
2011-05-13 Davide Pirino
* gbhill.c: Added option -O 5 for printing # of obs used to
stdout.
2011-05-11 Giulio Bottazzi
* gbplot: added option -l to set log scale on the axes and option
-C to issue generic commands before the plot
2011-03-24 Davide Pirino
* gbnlreg.c: Added option -O 4 to compute s-scores.
2011-03-10 Giulio Bottazzi
* gbbin.c: added option -w that allows for forced setting of the
boundaries. When this option is set, the bins are defined
accordingly to an uniformly spaced grid in the provided
interval. Notice that using this option leads to bins that are no
longer equally populated. The new option -w is incompatible with
options -x,-y and -c.
2011-02-28 Davide Pirino
* gbhill: warnings now only in verbose mode for better
visibility when gbhill is called repeatedly in scripts.
2011-02-25 Giulio Bottazzi
* Makefile.am: added script gbkeys2values
2011-02-11 Giulio Bottazzi
* gbget.c: multiple instances of option -t are joined together.
2011-01-19 Davide Pirino
* gbhill: added varcovar options for upper tail estimation for
gaussian, pareto1 and exponential. Note that pareto3 is still
missing and lower tail estimations are not implemented. Probably
they should be both removed from the code.
2010-12-11 Giulio Bottazzi
* gbplot: use 'wxt' as default terminal if available, otherwise
use 'x11'
2010-12-10 Giulio Bottazzi
* gbinterp.c: if number of points and range are not provided,
print 10 equally spaced points
* gbnlpolyit.c, gbtest.c, gbnlqreg.c, gbnlprobit.c, gbnlpanel.c,
gbnlreg.c, gbglreg.c, gbhisto.c, gbhill.c, gbgcorr.c, gbbin.c,
gbfilternear.c: moved 'initialize_program' after command line
processing
2010-12-03 Federico Tamagni
* gbhill.c: renamed estimated parameters, according to new
functional specifications adopted in exponential_gbhill.c and
paretoI_gbhill.c
* exponential_gbhill.c, paretoI_gbhill.c: new functional
specifications for distribution functions, F[j], and density
functions, -log(f[j])
2010-12-03 Giulio Bottazzi
* tools.h, tools.c: removed function "moment_gbbin" because no
longer necessary.
* gbbin.c: implemented the "xmedian" output option. Now both
xmedian and ymedian are computed only when needed (the ordering
operation implied in the computation of the median is N*log(N) in
time and better avoided if unnecessary).
2010-12-02 Giulio Bottazzi
* Makefile.am: added 'multimin.h' and 'gbhill.h' among the
distributed files
2010-10-11 Giulio Bottazzi
* gbnlprobit.c: updated help message with -g default value
2010-09-10 Davide Pirino
* tools.c, tools.h: added new function moment_gbbin called in gbbin.c
* gbbin.c: new option "ymedian" is now available
2010-09-03 Giulio Bottazzi
* gbhisto.c: added new option '-w' to manually set the binning
window
2010-08-22 Giulio Bottazzi
* gbkreg.c: corrected bug, added explicit cast to double type in
the computation of kernel values in method 1 (convolution).
* gbkreg.c, gbker.c: new boundaries for convolution
summation (method 1) should be more readable.
2010-08-21 Giulio Bottazzi
* gbkreg.c: corrected bug, added explicit cast to int type in
boundary definition of method 2.
2010-08-20 Giulio Bottazzi
* tools.c: the function 'load2' and 'load3' have been rewritten as
direct calls to function 'readblock', skipping the intermediate
call to function 'loadtable'.
* gbkreg.c, gbkreg2d.c, gbker2d.c: corrected bug in data loading:
the variable 'size' was not explicitly initialized to zero. This
is now necessary because of the new input functions.
2010-07-29 Giulio Bottazzi
* tools.c (loadblocks,loadblocks_bin): now these functions append
read blocks to an already existing block list (if block>0).
* gbget.c: explicitly set internal data storage to NULL after
freeing.
2010-07-22 Giulio Bottazzi
* gbnlpolyit.c: added computation of marginal effects in two ways:
total and elasticities. Added option -V to compute standard
deviations.
2010-07-20 Giulio Bottazzi
* multimin.c: added warning when the maximum number of iterations
is reached without finding a solution.
* gbnlpolyit.c: Added output to plot class-occupancies and
chi-square. Added log-likelihood among the output
2010-07-07 Giulio Bottazzi
* gbnlpolyit.c: added the new utility to obtain ML estimation of
the "polyit" model. Just a draft, it lacks several features.
2010-07-06 Giulio Bottazzi
* README: updated the description of 'sequential' data reading and
the fact that only the first block of data files is read.
* tools.c: load and loadtable functions have been rewritten to use
the readblock function. Now these functions only read the first
block.
2010-07-05 Giulio Bottazzi
* tools.c, tools.h: revision of data loading functions. New
version of binary function dropping unnecessary parameters. More
general management of blocks structure, in both ASCII and binary
functions. N.B.: now load_bin and loadtable_bin IGNORE the blocks
after the first. This is different from the behaviour of their
ASCII version.
* gbget.c, gbfun.c, gbmstat.c: updated management of binary files
input
2010-05-22 Giulio Bottazzi
* gbconvtable: added script to make substitution on a given column
of standard input given key-value dictionary file. Useful to group
data.
2010-05-20 Giulio Bottazzi
* gbget.c: remove unecessary creation of new matrices in
data_applyfuns and data_applyselection. Added function
data_applyfuns_recursive to implement recursively generic
transformation on data. The recursive interpretation is invoked
when the functions string begins with '@'.
2010-05-18 Giulio Bottazzi
* gbnlreg.c (main): print the number of observations in verbose
output
2010-05-03 Giulio Bottazzi
* gbdummyfy: corrected the use of '>' and '<' operator with -gt
and -lt respectively. These are the appropriate operators for
integer comparison.
2010-04-30 Giulio Bottazzi
* gbdummyfy: the new version based on gawk provides much better
performances.
2010-04-29 Giulio Bottazzi
* gbget.c (main): debugged nasty bug affecting the interaction
between option -t and transformation 'P'
2010-02-27 Giulio Bottazzi
* gbstat.c (main): fixed inconsistencies in help message
2010-02-22 Giulio Bottazzi
* gbstat.c: implemented option -O to select which statistics to
print. Notice that the output format has slightly changed (no more
dots but spaces and right indentation).
2010-02-19 Giulio Bottazzi
* tools.c (printnomatrixbycols): fixed bug which made this
function always print to standard output.
* gbget.c (data_applyselection): changed the way in which checks
are handled. Now comma separated conditions inside a couple of
curly brackets are OR-ed, while of course the specification of
several bracket couples serve as an AND specification.
2010-02-18 Giulio Bottazzi
* gbnlqreg.c (main): fixed initialization condition of simplex
size; added switch -M to select between minimization algorithms;
added option -s to scale initial simplex size; added option -N to
set the max number of iterations.
2010-02-16 Giulio Bottazzi
* gbget.c (data_applyselection): changed check implementation to
make it consistent with NAN
2010-02-03 Giulio Bottazzi
* gbget.c : added new (undocumented and to be tested) feature to
select rows according to rules: in the transformation space, specs
like '{fun}' apply fun to the row and if the result is greater
equal to zero print the line. More specs can be separated by
comma, like '{fun1,fun2,...}'.
2009-12-17 Giulio Bottazzi
* tools.c: used size_t to count the blocks in loadblocks(). In
printnomatrixbycols_bin() and printmatrixbycols_bin() added check
that the amount of required output is actually printed.
2009-12-12 Giulio Bottazzi
* tools.c, gbbin.c, gbtest.c, gbnear.c ,gbker2d.c, gbkreg2d.c,
gbboot.c, gbgcorr.c, gbfilternear.c, gbmstat.c, gbker.c,
gbkreg.c,gbinterp.c,gbfun.c,gblreg.c,gbglreg.c,gbnlreg.c,
gbnlpanel.c,gbnlqreg.c,gbrand.c,gbhill.c,gbnlmult.c, multimin.c:
fixed 'printf' spec for size_t output
* gbget.c, tools.c, tools.h: added a custom version of strndup for
systems without it (most notably MAC OS). The code is courtesy of
Stefan Soucek ssoucek at coactive.com
2009-09-11 Davide Pirino
* gbnlpanel: A isnan check was missing in the option -V1 -M0.
Help message has been updated.
2009-09-26 Davide Pirino
* gbnlpanel: Minor corrections.
2009-08-26 Davide Pirino
* gbnlpanel: Important NaN handling improvement: the program was
faulty for data where an entire realization should be removed
because of Nan's. Help message has been updated.
2009-08-26 Giulio Bottazzi
* gbdummyfy: little modification to error messages
2009-08-25 Giulio Bottazzi
* gbdummyfy: Minor corrections to help message.
2009-08-20 Giulio Bottazzi
* README: added mention of the GSL error handling behaviour.
* gbnlprobit.c: added option -O 6 to compute marginal effect using
the average of the derivative rather than the derivative computed
in the average. Notice that -O 5 is not implemented as should
become option 4 with error estimates.
2009-08-19 Giulio Bottazzi
* tools.c, tools.h, gbenv.c: GSL error handling is switched off by
the function 'initialize_program' if the variable
'GB_ERROR_HANDLER_OFF' is set in the environment. This requires
the insertion of a GSL header file in 'tools.h'. Hope this does
not endanger portability. The status of the handler can be
obtained using the 'gbenv' function.
* gbnlprobit.c: the verbose option prints also the value of the
score function at the optimal threshold.
2009-08-18 Giulio Bottazzi
* gbnlprobit.c: added colons in verbose output; implemented
output of marginal effects as option -O 4.
2009-08-04 Giulio Bottazzi
* multimin.c: verbose output of function and gradients are now
printed using scientific notation. The initial value of the
function is also printed.
* gbnlprobit.c: re-implemented the object function and derivatives
using the complementary error function and its log via the GSL
function 'gsl_sf_log_erfc'. Set the default minimizer to
BFGS2. Removed the computation of probabilities from the object
function. Modified variance-covariance matrix to use complementary
error function.
2009-08-03 Giulio Bottazzi
* gbnlprobit.c: improved the management of threshold
computation. Now by default any statistics based on threshold
levels is NOT computed. One has to explicitly ask for it using
option '-t'.
* multimin.c: print iteration number in verbose mode.
2009-07-30 Giulio Bottazzi
* gbnlprobit.c: remove z-scoring of constant variates.
2009-07-24 Giulio Bottazzi
* gbnlprobit.c: fixed wrong deallocation of var-covar
matrix. Fixed the code to properly recognize constant columns as
non-dummy.
* Makefile.am (EXTRA_SCRIPTS): added script 'gbdummyfy' to the
package
2009-07-23 Giulio Bottazzi
* multimin.h: added gsl_blas.h and gsl_linalg.h header files.
* gbnlprobit.c: build covariance matrix only if output option is
greater than 1. Build Brier score and threshold values only if
required.
2009-07-21 Giulio Bottazzi
* gbnlprobit.c: verbose messages properly redirected to stderr.
2009-07-20 Giulio Bottazzi
* gbnlprobit.c: added printing of the number of observations in
verobose output. Added warning for the case of completely
separated dummy variables (troublesome in linear models).
2009-07-16 Giulio Bottazzi
* gbboot.c: added output method -O 5 to bootstrap rows maintaining
the block structure
* gbplot: replaced 'return' with 'exit'
2009-07-15 Davide Pirino
* gbnlpanel.c: Computation of the variance-covariance has been
simplified and fitted to the latest version of BODIPI. Standard
output of panel statistics has been improved. Other minor
beautifications.
2009-07-09 Davide Pirino
* gbnlpanel.c: Final implementation of the variance-covariance
matrix estimation.
2009-07-02 Davide Pirino
* gbnlpanel.c: Corrections on the fixed effect case.
"denan" procedure totally implemented for the fixed
effect case.
2009-06-28 Giulio Bottazzi
* gbhill.c: added option -a to consider entire data-set when
printing distribution or density
2009-06-24 Davide Pirino
* gbnlpanel.c: implemented new variance-covariance matrix based on
last modification and computed on non-reduced log-likelihood.
2009-06-18 Giulio Bottazzi
* gbnlpanel.c: implemented new object function with prescription
of zero mean for fixed effect constant. Variance-covariance
expression must be re-worked.
2009-06-17 Davide Pirino
* multimin.c: minor beautifications.
* gbnlpanel.c: data structure has been modified: blocks identify
variables, row identify realizations, columns identify time. The
output possibilities now contain fixed or random effects
statistics.
2009-06-11 Giulio Bottazzi
* multimin.c: added Vector Broyden-Fletcher-Goldfarb-Shanno
version 2 method. Added printing of gradient modulus in verbose
output.
* gbnlreg.c: switched to the use of "multimin" interface. Output
improvement but still work to do.
2009-06-09 Giulio Bottazzi
* README: included extended description of the new 'gbplot'
command.
* Makefile.am (EXTRA_SCRIPTS): added 'gbplot' among the package's
scripts. This version supersedes the original examples provided in
the README file.
2009-05-30 Giulio Bottazzi
* gbstat.c: corrected bug in the printing of headings with the -t
option
2009-05-22 Giulio Bottazzi
* tools.c, tools.h : replace FILE with GBFILE to maintain
compatibility with compressed files in loadtableblocks and
readblock.
* gbnlpanel.c: Change name from " loadtableblocks_gb" to
"loadtableblocks". Added copyright notice.
2009-05-19 Davide Pirino
* gbnlpanel.c: Random effect has been implemented.
2009-05-11 Davide Pirino
* tools.c, tools.h: added functions loadtableblocks_gb and
readblock for acquiring panel data.
* gbnlpanel.c: First draft of nonlinear panel estimation. Only
fixed effect available.
2009-04-07 Giulio Bottazzi
* tools.c, tools.h: modified function mystrtod to accept a single
string pointer. Check for NAN is performed inside the function.
* tools.c, gbfun.c: modified calls to mystrtod.
2009-02-17 Giulio Bottazzi
* gbfun.c: added binary file management.
* tools.c: modified moment_nonan and moment_short_nonan to store
the number of finite entries in a variable of type double. Added
proper handling of output format in printmatrixbyrows() and
printmatrixbycols(). Added function printnomatrixbycols() to print
data which are not in matrix form. The "binary" IO method has been
implemented with the function printmatrixbycols_bin() and
printnomatrixbycols_bin() to print binary data and the functions
load_bin(), loadtable_bin() and loadtable_block_bin() to read
them.
* gbstat.c: moved initialize_program() after options parsing.
* gbmstat.c: there are two versions of this utility gbmstat_slow.c
and gbmstat_fast.c. The first has a straightforward design: a
slice of the input is passed to a routine which computes the
statistics. The second is carefully design to run faster: it
updates moments with new information so as to require less
computations. For some reason (?) their speed is comparable. So
I'll retain the slow version, which is easier to read, and keep
the fast for future development. Notice that the 'fast' version
does not have option '-D'.
2009-02-11 Giulio Bottazzi
* gbmstat.c: added new command to compute moving
statistics. Modified accordingly Makefile.am.
2009-02-09 Giulio Bottazzi
* gbmave.c: reverted to the old version. The modified new version
is now named gbmstat.c
2009-02-02 Giulio Bottazzi
* TODO: check the use of moment_nonan in gbstat and the policy for
the warnings about not enough observations.
* gbmave.c: new version with the ability to process column and to
compute different kinds of statistics.
* tools.c: added function 'moment_nan' to compute statistics
ignoring NAN entries and modified the existing function
'moment_short_nan' to provide, consistently with he previous one,
the number of not NAN entries. Notice that moment_nonan is silent
on the number of observation needed and simply returns NAN for a
variable if unable to compute it.
2009-01-30 Giulio Bottazzi
* gbtest.c: added test 'R' to compute Pearson's correlation and
p-score
* gbget.c: added transformation 'w' to normalize a column by
dividing each element by the sum of them. Barring NAN
entries,. the sum of the elements of the column is equal to one.
2008-10-22 Giulio Bottazzi
* gbquant.c: added option -e to print estimated error, using
asymptotic normal approximation, for the quantile value and
score (options -x and -1). Presently not implemented in table
mode (option -t).
2008-08-09 Giulio Bottazzi
* gbnlqreg.c: finally implemented computation of var-covar
matrix. The calculation is done considering the ML approach to
quantile estimation, i.e. starting from an asymmetric Laplace
distribution.
2008-08-07 Giulio Bottazzi
* gbquant.c: fixed bug: corrected the definition of min and max
indexes for the window selection explicitly using 'floor' and
'ceil' functions.
* gbnlqreg.c: corrected help output.
2008-07-23 Giulio Bottazzi
* gbget.c (main): updated help message
2008-07-22 Giulio Bottazzi
* gbget.c (new_matrix): fixed bug affecting the reversed count of
rows and columns when the skip is negative.
2008-07-12 Giulio Bottazzi
* scripts.txt: created a new file to collect useful scripts
* README: added mention of gbfilternear. Added description of shell
function 'gbploti'.
2008-06-04 Giulio Bottazzi
* gbbin.c: modified call to function sortn.
* gbquant.c: modified call to function sortn and added option -W.
* tools.c, tools.h: changed sortn to accept the position of the
column to use for sorting. Various modifications to the source
code, with the possible introduction of some bug.
2008-05-26 Giulio Bottazzi
* gbfun.c: fixed bug in the output of '-t' option.
2008-05-03 Giulio Bottazzi
* gbfilternear.c, Makefile.am: added program to filter near points
in Euclidean metrics. Useful to reduce the size of scatter plots.
2008-04-07 Giulio Bottazzi
* README: Added mention of the new programs.
2008-04-06 Giulio Bottazzi
* gbgrid.c: added option '-o' to set output format string.
2008-03-31 Giulio Bottazzi
* gbnlreg.c: explicitly freed some allocated space at the end.
* gbnlprobit.c: fixed bug in data storage allocation. Added option
-z to zscore non-dummy data. Added printing of Type I and II
errors. Added estimation based on score function and a long help
that could be possibly improved.
* tools.c: explicitly freed 'line' variable inside the data reading
functions.
2008-03-28 Giulio Bottazzi
* gbnlprobit.c: added computation of summary statistics. Output is
pretty ugly and should be improved.
2008-03-26 Giulio Bottazzi
* multimin.c (multimin):
* Makefile.am, configure.ac, gbnlprobit.c: added gbnlprobit to the
list of utilities.
* multimin.c: a NULL type array is equivalent to no transformation
on the variables. Simplex hedge size initialized to
step_size+maxsize.
2008-03-15 Giulio Bottazzi
* gbhisto2d.c: added options to handle discrete asymmetric data,
that is couples where the two elements belongs to different sets
of values.
2008-03-14 Giulio Bottazzi
* gbget.c: corrected nasty bug in transformation identification.
2008-03-13 Giulio Bottazzi
* ALL: present version becomes 5.3 (beta)
* gbfun.c: some code beautification.
* Makefile.am: added proper linking requirements for gbnlmult.
* README: updated README with new utilities and features.
* gbget.c: improved parsing of spec, now the order of the block
and slice specification can be reversed. Added new transformation
'<..>' that substitutes the matrix of selected data with a set of
function computed on it. Added separator ';' to specify different
slices of rows and columns.
2008-03-12 Giulio Bottazzi
* gbfun.c: added option '-o' and '-s' to set output format.
* gbget.c: created print-data function to handle all the
printing. The check to see if data are in matrix form is now
performed only there. Added transformation 'P' to print the data
collected so far followed by two empty lines. In this way it is
possible to implement a block-wise output. Now retrieval specs are
builded from the command line in an intelligent way. Improved
check on retrieval spec structure.
2008-01-30 Giulio Bottazzi
* tools.c, gbstat.c: corrected a mistake in "moment" function:
with one observation all the statistics are (almost) perfectly
defined. Updated the help message in gbstat.
2008-01-16 Giulio Bottazzi
* ALL: present version becomes 5-2-2 (tagged with CVS) using
#cvs tag gbutils-5-2-2
This can be recovered using
#cvs checkout -r gbutils-5-2-2
* Makefile.am (EXTRA_PROGRAMS): removed fake entry for "gbnlmulti"
* gbker.c: added option '-p' to print confidence interval (as an
extra column) computed according to eq. 3.51 in Hardle W., Muller
M., Sperlich S. and Werwatz A. "Nonparametric and Semiparametric
Methods" (2004) Springer-Verlag, Berlin. Fixed a bug in method
1-covolution, adding an explicit cast before taking differences of
variables of type size_t as double. Other bugs could in principle
be present.
2007-12-19 Giulio Bottazzi
* gbtest.c: corrected name of variables in verbose output
2007-11-29 Giulio Bottazzi
* gbquant.c: fixed the order in which the function data_denan is
applied when invoked with option "-w".
2007-08-14 Giulio Bottazzi
* multimin.c: fixed output of multimin for simplex initial
settings
2007-08-09 Giulio Bottazzi
* gbboot.c : added option -O 4 to have data "displaced" by a
fraction of the minimal non-zero distance between consecutive
observations.
2007-08-07 Giulio Bottazzi
* gbinterp.c: added option -I to read interpolated points from a file
2007-08-03 Giulio Bottazzi
* gbnlqreg.c: fixed the help message
* gbtest.c: added Spearman's rho and Kendall's tau for correlation
measure between two samples. The implemented statistics are valid
with or without ties and are defined according to W.J. Conover
"Practical Nonparametric Statistics", Third Edition.
2007-07-20 Giulio Bottazzi
* gbtest.c: added computation of Spearman's Rho correlation.
2007-05-10 Giulio Bottazzi
* gbutils.css: new CSS file.
2007-05-02 Giulio Bottazzi
* gbtest.c: corrected typos in help messages.
2007-04-25 Giulio Bottazzi
* gbnlqreg.c, configure.ac, makefile.am: added a new utility:
gbnlqreg
* gbnlreg.c: added mention of information matrix computation in
comments for var_covar option number 3.
2007-04-05 Federico Tamagni
* gbnlmult.c: fixing the help displayed to users.
2007-04-03 Giulio Bottazzi
*Makefile.am, configure.ac: inserted gbnlmult in the list of
utilities
* gbnlmult.c: added gbnlmult to fit systems of simultaneous
equation with OLS, that is assuming normally distributed shocks.
2007-03-30 Giulio Bottazzi
* ALL: present version becomes 5.2.1
* tools.c: updated copyright notice
* gbfun.c: added 'lag' operator to specify previous values. For
instance 'x3l4' stands for column 3, four steps before. Notice
that the first row of output is build accordingly, so if some lag
is specified, the output number of rows is reduced.
2007-03-27 Giulio Bottazzi
* ALL: --- RELEASED VERSION 5.2 -------------------------------
* gbgcorr.c: added a short description of the program
* gbutils.css: added a stylesheet file to generate html
documentation
* README: added mention of new files in the list. The html is
generated with
rst2html.py --stylesheet gbutils.css --embed-stylesheet
2007-03-26 Giulio Bottazzi
* gbenv.c : removing output of floating point environment under
Cygwin, where the ieeefp.h seems not to work.
2007-03-25 Giulio Bottazzi
* gbenv.c (main): added floating point environment reporting for
Cygwin, which is based on ieeefp.h not present under Linux.
2007-03-23 Giulio Bottazzi
* tools.c: correctly named "THOUSEP" the thousand separator instead
of THOUSANDS_SEP.
* Makefile.am : added necessary source files to gbhill
2007-03-13 Giulio Bottazzi
* README: added mention of "gbenv" and "gbhill" in the listing of
programs and in the table of properties.
2007-02-09 Giulio Bottazzi
* gbboot.c: corrected mention of default sample size in gbboot
help.
2007-02-08 Giulio Bottazzi
* gbhill.c (main): corrected bug: added check of parameters
number and the loading of their values in the pareto1 case.
* README, README.gbutils, INSTALL.CYGWIN: with the most recent
version of docutils (ver.0.4) simply use
rst2html.py --embed-stylesheet
* README: added a brief section on locale management
2007-01-23 Giulio Bottazzi
* gbfun.c: changed strtod instances to mystrtod.
* tools.c, tools.h: added function mystrod to properly parse
thousands separator defined by the locale. It only works with
single-character separators like ',' or '.' This function uses
the external variable LCTHSEP which is initialized with the
setting in the environment.
2007-01-22 Giulio Bottazzi
* tools.c, gbenv.c: added locale initialization from environment
variables through function "setlocale".
2006-12-13 Federico Tamagni
* gbhill.c: fixed bugs in initial values for pareto1 pareto3 and
gaussian fitting
2006-12-05 Federico Tamagni
* gbhill.c: added a switch to check initial parameter values.
* paretoI_gbhill.c, paretoIII_gbhill.c, gaussian_gbhill.c: added
two -M options to compute estimates conditioning on threshold
levels. Notice that now -M 0,1,2 refer to upper tail inference,
while -M 3,4,5 to lower tail estimation.
2006-11-28 Giulio Bottazzi
* gbhill.c, exponential_gbhill.c: added two -M options to compute
estimates conditioning on threshold levels. Notice that now -M
0,1,2 refer to upper tail inference, while -M 3,4,5 to lower
estimation.
2006-11-22 Giulio Bottazzi
* gbrand.c: Added Pareto type III distribution.
2006-11-20 Giulio Bottazzi
* gaussian_gbhill.c: commented out printing of debug output.
2006-11-17 Giulio Bottazzi
* gaussian_gbhill.c: fixed some mistakes in derivatives
expressions.
2006-11-11 Federico Tamagni
* gbhill.c: removed bug in initialization of s parameter for
pareto3 fitting.
2006-11-10 Giulio Bottazzi
* gbtest.c: fixed bug in expected variance of TR-RT test, the
formula in Brockwell and Davis is actually wrong.
2006-11-10 Federico Tamagni
* gbhill.c: removed bug in initialization of s parameter for
gaussian fitting.
2006-11-09 Giulio Bottazzi
* gbtest.c: fixed typos in doc strings.
2006-11-08 Giulio Bottazzi
* multimin.c: fixed typos in doc strings.
* gbhill.c: removed old "print true values" code. It was a remnant
of a previous implementation. Added minimum value of the negative
log-likelihood to the output. Added switch to properly select
minimization method and boundaries for the different cases of the
exponential.
2006-11-08 Federico Tamagni
* gbhill.c: removed bug in initialization of b parameter for
exponential fitting.
2006-11-06 Giulio Bottazzi
* gbhill.c: replaced minimization type 5 with 2 in
estimating exponential's b parameter; remove bug in initialization
of k value for exponential. Replaced minimization type 1 with 4 in
estimating Pareto's b parameter.
2006-10-17 Giulio Bottazzi
* gbgcorr.c: added Cees Diks' gbgcorr to compute Gaussian kernel
correlation dimension of time series.
2006-10-14 Giulio Bottazzi
* ALL: present version becomes 5.1.4
2006-09-28 Giulio Bottazzi
* gbenv.c : added display of locales information
2006-09-26 Giulio Bottazzi
* gbhisto2d.c: added management of discrete variables; notice that
discrete states can also be labeled by floating point
numbers. All the computation has been integrated in the main body
of the program.
* gbhisto.c: improved some error message and general outline.
* gbtest.c: improved some error messages.
* tools.c: added function my_calloc to allocate memory initially
set to zero;
2006-09-22 Giulio Bottazzi
* gbtest.c: added Levene test for equality of variances in 2 or
more groups, robust to non-normality, see Brown, M. B. and
Forsythe, A. B. 1974, Journal of the American Statistical
Association, 69, 364-367. Made the output of F test more
verbose. Changed the way in which the name of the test is printed.
* gbstat.c: fixed bug in the determination of the median with
option "-t".
2006-09-18 Giulio Bottazzi
* gbquant.c : enhanced option -w to handle multi-columns input;
modified help description.
* tools.c: moved here from gbget.c the definitions of "data_denan"
which remove entire records, that is rows, containing NAN entries.
* README : updated the table with the properties of the utilities
2006-09-15 Giulio Bottazzi
* gbbin.c : added outtype=NULL in the -O option; before passing it
to realloc, a pointer not initialized should be set to NULL.
2006-09-01 Giulio Bottazzi
* README: changed table entry relative to the way in which gbbin
handles input data: from sequential to tabular. Updated version to
5.1. TODO: add mention of gbhill.
* gbbin.c: changed behavior of gbbin to handle multiple columns
of y entries. Added options -c to split the elements of a list of
columns.
2006-08-10 Giulio Bottazzi
* tools.c : added function sortn to sort all the columns of a
table with respect to the values on the first column.
2006-05-25 Federico Tamagni
* paretoIII_gbhill.c: fixed bugs in computation of likelihood
derivatives.
2006-05-25 Giulio Bottazzi
* paretoIII_gbhill.c: fixed bugs. Probably some error still
remain.
* gbhill.c: added more verbose warnings; constrained density and
distribution output (-O 1 and -O 2) to the used observations;
fixed pareto3 initialization..
2006-05-25 Federico Tamagni
* paretoIII_gbhill.c: file added for Pareto Type III tail fitting with
gbhill. Still a draft.
* gbhill.h, gbhill.c: files have been modified to include
Pareto Type III fitting.
2006-05-18 Giulio Bottazzi
* gbhill.c: added output option '3' to print Renyi-transformed
residuals.
2006-05-06 Giulio Bottazzi
* gbtest.c: added three tests of randomness: TR-TP, TR-DS and
TR-RT. Added exact and asymptotic distribution under the null
hypothesis for D+ and D- tests. The name of Kolmogorov-Smirnov
statistics has been changed to D.
2006-04-26 Giulio Bottazzi
* gaussian_gbhill.c: fixed the expression for the derivative dFdm.
2006-04-25 Giulio Bottazzi
* gbhill.c : minor beautifications.
2006-04-22 Federico Tamagni
* gaussian_gbhill.c: fixed some errors in computation of
likelihood's derivatives. Estimation using all the data seems ok,
but that using a fraction of the sample still have problems.
* gbhill.c: changes made in the initialization of the parameters
for the Gaussian case.
2006-04-20 Federico Tamagni
* gaussian_gbhill.c: file added for gaussian tail fitting with
gbhill. Very preliminary, estimation has several problems.
* gbhill.h, gbhill.c: files have been modified to include
gaussian fitting.
2006-04-13 Giulio Bottazzi
* exponential_gbhill.c, paretoI_gbhill.c, gbhill.h: definitions of
different distributions have been split in different files to keep
gbhill.c smaller (and more readable).
* tools.h: defined GBTOOLS_H variable to load the tools.h header
file only once.
* gbhill.c: fixed error in computation of Pareto Type I
derivatives. Case M=3 seems to have problems.
2006-04-12 Federico Tamagni
* gbhill.c: added Pareto Type I fitting, fixing boundaries of
minimization for this case. Very preliminary, minimization has
problems.
2006-04-06 Giulio Bottazzi
* gbhill.c: fixed estimation when k=N. New version with all four
estimators, upper and lower, conditional and unconditional.
2006-04-04 Giulio Bottazzi
* gbhill.c: new version of gbhill; still a draft.
2006-03-31 Giulio Bottazzi
* gbnlreg.c: fixed typo.
* gbtest.c: added check that data are from cumulated
distribution(s) for the 1 sample Kolmogorov-Smirnov or Cramer-von
Mises tests. Added verbose output to KS test.
2006-03-29 Giulio Bottazzi
* gbtest.c: added KS two samples to the displayed list of tests.
2006-03-24 Giulio Bottazzi
* gbhill.c: fixed boundaries of minimization in exponential
fitting.
2006-03-22 Giulio Bottazzi
* tools.h: fixed path of .h includes from lib/ subdirectory.
* gbhill.c: implementation of exponential fitting seems
ok. The utility is still a draft.
* multimin.c: a bit of beautification.
2006-03-21 Giulio Bottazzi
* ALL: present version becomes 5.1.3
* gbhill.c: added new program for the estimation of tail
behaviour. Just a draft.
2006-03-20 Giulio Bottazzi
* gbker.c: fixed bug in the initialization of variable "imin" in
method 2, exact summation.
2006-03-16 Giulio Bottazzi
* gbtest.c, gbenv.c, gbrand.c: removed typos and non-dangerous
bugs.
2006-03-14 Giulio Bottazzi
* gbtest.c : added Student's T test for single sample.
* gbmave.c : changed option "-t" in "-s". Previous name conflicted
with general usage of "-t" as indicating "tabular" behavior.
2006-03-13 Giulio Bottazzi
* ALL: present version becomes 5.1.2
* gbget.c: added check at the end to skip printing if no data were
loaded.
* gbtest.c: removed option "-d" no longer used. Improved some
output with option "-v 2". Tested CHI^2, Wilcoxon, Mann-Whitney,
Kruskal-Wallis and Fligner-Policello. Added F-test and Student's T
test in the homoscedastic and heteroscedastic cases.
2006-03-10 Giulio Bottazzi
* gbtest.c: added chi^2 test for 1, 2 and 3+ samples, that is
contingency table analysis. Added more verbose option "-v 2"
which prints detailed output.
2006-03-08 Giulio Bottazzi
* tools.c: fixed bug in the loading functions, like load,
loadtable, etc. affecting the parsing of the last line of the
input file which does not end with a '\n'.
* gbtest.c: added Wilcoxon and Fligner-Policello tests, and the
relative large-sample p-scores, for both 1 and 2 sided cases.
* gbrand.c (main): added "discrete" trasformation to generate
integer from 1 to M with generic probabilities. Probabilities are
passed as comma separated list of values.
2006-03-05 Giulio Bottazzi
* gbtest.c: added Kruskal-Wsllis multi-sample test and
p-score. The latter introduced a dependence of gbtest on gsl which
has been added to the table in README.
2006-03-04 Giulio Bottazzi
* gbtest.c: fixed bug in run_onesample when pscore=NULL. Added
function run_manysamples for tests which uses multiple
samples. Added specifier const to data pointer used by
run_... functions.
2006-03-01 Giulio Bottazzi
* gbenv.c: added utility to print floating point and gbutils
settings.
2006-02-28 Giulio Bottazzi
* gbnlreg.c: reverted the stopping condition of the estimation
loop of the simplex algorithm to simplex size-check. This is the
only consistent method. Slightly modified verbose output. Print
numerical precision at the beginning of the iterations (if verbose
enough).
2006-02-27 Giulio Bottazzi
* Makefile.am: "m4/CVS" subdir removed from distribution tar.gz
* ALL: present version becomes 5.1.1
* gbnlreg.c: fixed bug related to stopping criteria: assignment
instead of comparison in "estimate_status == GSL_CONTINUE".
2006-02-09 Giulio Bottazzi
* gbnlreg.c: modified the stopping condition for the estimation
loop. Now the reduction, absolute and relative, of both the object
function and the parameter estimations are considered. The aim is
to have a stopping condition which scales well with the increase
of the number of observations.
2006-02-08 Giulio Bottazzi
* gbnlreg.c: added option '-e' to set the minimization
tolerance. The default tolerance has been decreased to 1e-5.
2005-12-13 Giulio Bottazzi
* gbtest.c: added pscore function for D (one-sample
Kolmogorov-Smirnov) test; corrected bug in run_onesample function.
2005-12-12 Giulio Bottazzi
* README : gbget added to the list of NAN-transparent utilities
* gbget.c: modified options z and Z to ignore NAN entries in the
computation of mean and averages
2005-12-10 Giulio Bottazzi
* gbget.c: added option "-t" to gbget to define FINAL
transformations; improved output management: removed unnecessary
checks when output has a matrix structure; added check at the
beginning of data_denan to see if NAN values are actually present.
2005-12-04 Giulio Bottazzi
* ALL: --- RELEASED VERSION 5.1 -------------------------------
* gbbin.c (main): now statistics are printed following the order
in which they are provided with the option -O.
2005-12-03 Giulio Bottazzi