From the 1.7 to the 1.76 version

(2018/12/10): Grocer version 1.76 has been posted on the web.
  • A new function deming, along with is lower level counterpart deming1, its printing function prt_deming and its demo function deming_d have been created. Corresponding help file has been also created.
  • The code of many functions manipulating ts and tsmat has been modified to be able to deal with ts and tsmat created with different representations of sub-annual frequencies (for instance 4 or [4,1] for quarterly frequencies).
  • Function brybos has been modified to better deal with annual data. If you have an annual series x test for instance:
    --> brybos('x','M=2','m=2','e=1','k=2');
  • multivariate test functions arlm0_multi and hetero_sq0_multi have been adapted to Markov-switching results; for instance save the result of a ms_var and run:
    --> result=ms_var([your input]); [val,pvalue]=arlm_sq0(result,4) ;[val,pvalue]=hetero_sq0(result) (for each test, val is the value of the test statistics and pvalue its p-value).

    (2018/06/19): Grocer version 1.75 has been posted on the web.

  • The code of several functions has been slightly modified to avoid uncessary warnings with Scilab 6.0.1.
  • Bugs in gmm and iv function, that did not work with ts, have been corrected.
  • The user can now enter a time trend in any econometric functions through the string 'trend' as well as through the the string 'trend^1'.
  • A new function addts2tsmat, that adds a ts to a tsmat, has been added.

    • (2018/02/12): Grocer version 1.74 has been posted on the web.

  • Panel estimation functions pppoled, pfixed and parandom have been adapted to work with unbalanced panels.
  • The automatic programm, that selects automatically the "best" model from a selection of variables, has been extended to deal with panel ppoled, fixed and random effects (see help automatic or have a look at the demo panelauto_d.sci for examples)
  • The disaggregation method that is used at Insee for quarterly national accounts has been tested, corrected and documented (see help etalcalinsee or have a look at the demo etalcalinsee_d.sci for examples).
  • Function expbd2exc allows the user to give different names to the variables in the destination file from their Scilab ones (see help expbd2exc).
  • Function reliability has been extended to work with panel and restricted VAR estimation methods.
  • Calculation of the log-likelihood in function sur, performing seemingly unrelated regressions, has been corrected.
  • Functions plt_garch, pltdma_expnbvar, pltdma_expnbvar and pltfac_kalm (plotting the results of garch, DMA and fac_kalman results) did not work with ts: this has been corrected.
  • The low-level functions explon and explone, used by most high level econometric functions, have been corrected to perform proprerly the determination of bounds when they are determined by the functions.
  • Function nls has been improved to be able to deal with complex non linear equations.
  • The error message displayed by function q2a has been corrected.
  • Function %tsmat_mean (not documented), that was not working properly, has been corrected and now works properly.
  • Function varmaf, that makes forecasts from VARMA models, has been corrected to deal properly with exogenous defined implicitely, such as 'const' or 'trend^2'.

    • (2017/03/11): Grocer version 1.73 has been posted on the web.

  • This version contains new adaptations to the Scilab 6 family, entailed by the transition from Scilab-6.0.0.beta2 to Scilab-6.0.0 (in particular, all strings written on several lines were put on a lone one; some remaining additions of an empty matrix and a non empty one have been removed and replaced by lines of codes working with Scilab versions both before and after 6.0.0).
  • This version corrects a bug introduced when adapting the ms_var and ms_reg programs to Scilab-6.0.0.beta2: the programs did not work anymore when all coefficients were switching.
  • Function johansen did not always work with ts: this has been corrected.
  • Function readxls2bd has been extended to work with annual dates written as numbers instead of being suffixed by 'a'.
  • The program automatic worked only with one compulsory variable; now it works with any number of them.
  • The program irf that provides the impulse function from a VAR has been rewritten. The option 'meth=mc1" has been discarded and replaced by the more relevant option 'meth=bootstrap', that estimates the confidence bands by a bootstrap method. Function pltirf2 that plots the corresponding result has been improved.
  • A new option 'saturate_post', that determines the presence of breaks to the constant term in a pc-gets like regression has been introduced to function automatic.
  • Help files have been submitted to an extensive review: files that had not been updated to the more recent help files specifications have been upgraded; some errors and typos have been corected.

    • (2016/07/23): Grocer version 1.72 has been posted on the web.

    This version is the first one to work with the Scilab 6 family, whose 6.0.0.beta2 version has been released in June 2016. This has entailed the modification of an unusal number of functions. In particular, many functions have been modified to deal with a major change with the Scilab 6 family: the result of the addition or subtraction of an empty matrix and a non empty matrix has changed, from the added non empty matrix to an empty matrix. I have performed extensive tests to discover all cases where it may happen in Grocer programs, but I cannot guarantee that there remains none. So if you have the message "Warning adding a matrix with the empty matrix will give an empty matrix result", do not hesitate to send me an e-mail, so that I can deal with the potential implied problems. The installation device has also been changed for Scilab 6 family, due to a change in the working in the lib function.

    This version corrects also a few bugs:

  • Option 'dropna' used to ignore the NA values in a daily ts now works properly on functions da2q and da2m;
  • When used with a second argument to restrict the names of the variables found in the database, function dblist retained erroneously some variables, now it gives the exact list of variables respecting the entered restrictions;
  • Function johansen_common_beta, that worked only in the very particular cases contained in the demo for the johansen suite, now works in the general cases;
  • Function prt_multilogit, that did not search for the coefficients values in the results tlist, now does.
  • Functions predfailin0 and auto_waldf0 sometimes generated an error because of the switching of a real number to a complex one with zero imaginary part: this has been corrected;

    • A few functions have also been improved:

  • Speed of function deriv_eq1 has been greatly increased. The working of function analyse_eq has also been changed: it now produces a global variable called grocer_variables, that collects the names of all variables found during the analysis;
  • Function values has been extended to allow the extraction of only a subset of variables in a tsmat;
  • Function %ts_mean %tsmat_mean has been created to overload function mean with ts and tsmat typed data.

    • (2015/11/09): Grocer version 1.71 has been posted on the web.

    The 1.71 version is a minor evolution of the 1.7 version, with the following changes:

  • Function impexc2bd has been extended to allow the exportation of comments in time series and to suppress the 'a' in Grocer year dates;
  • Erroneous } signs in functions IPS1 and select_mask have been replaced by the correct ] sign;
  • In functions ms_reg and ms_var, the possibility of entering options to optim has been introduced.
  • There was a bug with the symbolic derivative with respect to the log function, that was not properly calculated when the expression between brackets contained operations.

    (2015/10/03): Grocer version 1.7 has been posted on the web.

    The main novelties of the 1.7 with respect to the previous 1.67 version are the following:

  • Function automatic has been extended to incorporate the block search algorithm proposed by J. Doornik to deal with problems when the number of variables is greater than the number of observations. Grocer block search has been compared to Doornik Autometrics block search program for the battery of tests used by Doorik in his 2009 paper. These results of these tests are presented in chapter 15 of Grocer manual and can be reproduced using the material given here: they show that Grocer block search program compares favourably with Autometrics one.

  • In the same vein, function ols has also been extended to allow impulse saturation, the methodolgy proposed by Castle, Doornik and Hendry to test the the presence of structural breaks in an equation.

  • A new Function Hetero_TVP_VAR has been created to allow the estimation of a structural VAR with varying coefficents and volatilities, estimated in bayesian framework using a Monte-Carlo Markov Chain (MCMC) algorithm. The user can adapt the priors to her needs and a function tvpvar_prior helps helps her to enter her choices. The function also come with a function Hetero_TVP_VAR_stresid that plots the -time varying- standard deviations of the residuals and a function Hetero_TVP_VAR_irf that plots the -time varying- impulse response functions of the VAR. It also comes with a function tvpvar_diagnos that performs convergence diagnostics. Latsly, functions MCMC_autocorr, that calculates autocorrelation of draws for a given set of parameters, geweke_diagnos that performs Geweke''s numerical std error, relative numerical efficiencies and inefficiencies and raftery that performs Raftery and Lewis (1991) convergence diagnostic, that is the # of draws needed in MCMC to estimate the posterior cdf of the q-quantile to within +/- r accuracy with probability s, can be used for various other bayesian estimations using the MCMC algorithm. Function wish_rnd and wish_rnd1, that makes randow draw in a Wishart law have also been added.

  • A function msvar_irf_cb, written by Stefan Fiesel, that performs confidence bands around impulse response functions in a Markov-switching VAR model, has been added. This has also lead to write a function msvar_draw that draws at random an artifical Markov-switching VAR with any set of parameters.

  • Functions findobject and strsubst_trueobj, that find if an object name is truly in a mathematical formula and to replace object names by another name in a mathemtical formula, have been documented.

  • Minor adjustements have been made to numerous functions, in order to adapt them to the coming Scilab 6.0 version: this should not change Grocer working for older versions, but has nevertheless the advantage to make the code somehow cleaner (Grocer cannot be installed for the moment with the Scilab 6.0.alpha, because the installation programs has yet to be adpated; but Grocer functions can nevertheless be run using less straightforward means: if you are an experienced user and are interested in testing Grocer under Scilab 6.0.alpha, send me an a-mail and I will indicate you how to proceed).

    • The user manual has been modified accordingly and slightly enhanced (correction of typos, of a few errors,...).

    From the 1.6 to the 1.67 version

    (2015/05/25): Grocer version 1.67 has been posted on the web.

    Changes to the 1.66 version are the following:

    - Working of functions optimg and nelmead has been slightly improved: the slowest call to optimg has been put at the end of the optimization loop and nelmead now gives a warning message when the maximum number of iterations has been reached
    - function histo_ts has been corrected to avoid side effects (not present in Scilab 5.4 family, but appeared with the 5.5 one) when called in a subplot
    - a new option 'win=x' has been added to function pltuniv has been added to allow the user to set the graphic window to number x
    - function invxpx has been changed to improve the precision of the estimated variances of coefficients in most econometric functions
    - functions explon, explone and deal_varspec have been improved to deal corretcly the cases when the user has given no bounds and there are NA values in the series

    (2015/03/22): Grocer version 1.66 has been posted on the web.

    Changes to the 1.65 version are the following:

    - Upon a request by Brian Bouterse, functions to run Markov-switching models out-of-sample, ms_reg_oos, ms_var_oos and ms_mean_oos, have been added to Grocer, along with corresponding demos (functions ms_reg_oos_d, ms_var_oos_d and ms_mean_oos_d). Help files have been created accordingly (run -->help ms_var_oos -->help ms_reg_oos or -->help ms_mean_oos).
    - Function MSVAR_Vec_Prob has been slightly changed to avoid overflow in some rare cases: functions ms_reg, ms_var and ms_mean should better converge in these cases.
    - Function MSVAR_Setinit has been again changed to prevent it aborting because some calculations became complex.
    - Performing estimations with trends defined implicitely ('trend^2' for a squared trend for instance) worked only for ts and not for vectors or matrices: now this can be done in both cases.
    - Handling by functions explon, explouniv and explone of variables that can belong to the list of endogenous variables and, lagged, to the list of exogenous variables (such as in VAR models and their derivatives Markow-swithing VAR, johansen cointegration method...) has been improved and a new function ts_span has been created to deal with this. The functionning of any working program should not be altered, except when applied to ts without explicit bounds: the bound should now be calculated by these programs over a longer period.
    - A new option to function pltseries0 has been added, that allows to impose (in pltseries as well as in pltseries0) the lower end and/or the upper end of the y scale ('miny1=x' -resp. 'miny2=x'- if you want to impose the lhs -resp. rhs- y scale to start at x and the same working for the upper end of the y scales: 'maxy1=x' and 'maxy2=x').
    - Messages have been added to function optimg to make the optimization process more easy to monitor.
    - A few elseif, erroneously written elsif, have been corrected (in functions autocum and AndPlob).
    - A few double quotes that ended simple quotes, something that does not hurt released Scilab versions, but that Scilab 6 won't like, have been changed to simple ones (a very small proprotion of Grocer simple quotes)

    (2015/01/22): Grocer version 1.65 has been posted on the web.

    Changes to the 1.64 version are the following:

    - Function str2vec has been extended to allow, in particular, quotes in graph legends.
    - Calculation of the hessian by function hessian0 has been slightly improved (from a matlab program by Hang Qian). Functions performing turning points indicators (ms_quali and some of its subroutines) have been modified accordingly.
    - A new function clean_sheet, that removes empty lines and colums from the text and value of an Excel sheet, has been added and function readxls2bd has been modified accordingly.
    - Function msvar_irf did not work properly for some options: this is corrected; titles of corresponding graphs, provided by function pltmsirf1 have been made more epxlicit.
    - Minor improvements have been made to function deriv_eq providing the symbolic derivative of an equation.
    - Some functions have been changed to prevent them aborting because some calculations became complex: functions e4prees2 (used in some state-space models), MSVAR_Setinit used by all Markov switching programms
    - A correction has made to function nelmead to make the optimization function optimg work better.
    - A bug in function cholp has been corrected.
    - Non important changes have been made to function varma1 and e4prees1.

    (2014/10/09): Grocer version 1.64 has been posted on the web.

    - A few corrections have been made for Grocer to work properly with Scilab 5.5.1: function hetero_sq0 peforrming heteroskedastcity tests has been corrected to be able to deal with the case when there is exact colinearity between some of the regressors; various functions used by function AndPlobthe performing Andrews and Ploberger test have been changed to avoid unwelcome warning messages.
    - function acf1 has also been sliglty modified to avoid unwelcome warning messages.
    - function analyse_eq has been enhanced to be able to deal with '+-' sequences and function rebuild_eq to deal with empty input.
    - Small improvements have been made to functions probit_like and probit1 used by Probit estimation, to function nelmead used by optimg optimisation function and %tsmat_p that display tsmats.
    - Option 'short' in function dblist was not working properly: it now works.

    (2014/07/24): Grocer version 1.63 has been posted on the web.

    This version corrects a few bugs: the legends, that did not appear on graphs with the Scilab 5.5.0, now appear correctly; the installation on Scilab root (an instllation mode that I do not recommend, but that I will maitain since some users are acustomed to it) did not work with the more recent Scilab versions; this has been repaired and made simpler to maitain; within the automatic device, a bug in auto_stage0 (the index of chosen variable was wrong when all variables where significant) has been corrected and minor changes, transparent for the user, have been made to a few functions; prtuniv, prt_res and rolvar_d did not work with Scilab 5.5.0, and their proper functionning has been restored.

    3 New functions, adapted from Scilab lines of code, have been added: a function getconfig that returns the name of the O/S; a function internet2sci, that downloads a file from the Internet (vey similar to the new Scilab function curl, but gives an error status as an output); a function sci_unzip, that unzip a file from Scilab. Defaults have been changed in functions ms_mean, ms_reg and ms_var to achieve convergence more frequently (but recall that bad convergence can still happen and in some cases be dealt with a change in the opimisation parameters of the function, see chapter 6 of the user manual). The possibility to add a legend has been added to function fan_chart.Lastly garch function has been extended by Emmanuel Michaux so as to be able to deal with garch models without any variable in the first stage.

    (2014/04/26): Grocer version 1.62 has been posted on the web.

    This version does not contain new functionnalities, but contains the modifications needed to run properly on Scilab 5.5.0. Moreover, functions irf_asy and irf_mc1, that generated errors since the 1.6 version, have been corrected to work again porperly.

    (2014/03/11): Grocer version 1.61 has been posted on the web.

  • This new version contains new functions: vargranger that performs Granger causality tests in a VAR; rolvar that performs VAR estimations on rolling periods; rolirf that calculates Impulse Response Function for rolling VAR; rolvar_granger that performs Granger (non-)causality tests for a rolling VAR (run help vargranger, help rolvar, help rolirf and help rolvar_granger for details).

  • In function pltseries0, an option to define the size of the graphic window has been added and allows to enter less items in the legend than the corresponding numbers of series.

  • Several bugs to function automatic and automatic_signed, mainly relative to corner solutions, have been corrected (thanks to intensive testing for an empirical project).

  • All '^' applied to vectors have been replaced by '.^' to comply with requirements of future Scilab releases.

    • (2013/10/13): Grocer version 1.6 has been posted on the web. The main novelties of the 1.6 with respect to the previous 1.55 version are the following:

  • new panel functions to estimate panel models with common correlated effects (written by Emmanuel Michaux) have been introduced;

  • a function DMA performing dynamic model averaging, a bayesian estimation method that deals with uncertainty on the true variables among N ones, while allowing time-varying parameters on these variables, has been added;

  • a function varwithfac that allows to introduce factors in a standard VAR has also been added;

  • function doornhans, that performs Doonik and Hansen normality test, has been extended to deal with multivariate models and 2 new functions, arlm0_multi and hetero_sq0_multi, have been created to perform the LM autocorrelation test and the heteroskedasticty test on squared exogenous variables on multivariate models;

  • an optimization function optimg that mixes Scilab optim function and Grocer nelmeade function to gain robustness has been introduced; at the occasion, all functions using an optimization program have been modified to be able to switch between the optim and optimg optimization functions and to allow the user to enter any option to these optimizers; this new function passes all 27 problems provided by the National Institute of Standards and Technology (see http://www.itl.nist.gov/div898/strd/nls/nls_main.shtml)

  • a syntactic analyzer (function analyse_eq) and a symbolic derivative function (deriv_eq) have been added; function nls, that performs non linear least squares, now uses whenever possible the symbolic derivative instead of a numeric one;

  • function automatic, that aims at mimicking the LSE methodology, has been improved along several dimensions: first, for more efficiency, specification tests are now performed at the end of a "path", and no more in a systematic way, and, when specification test are not accepted, then the model is backtracked until they are; second, with a new option 'depth=n' the user can now test more models than previously; and third, the function has been adapted to deal with Probit and VAR models, and can be extended by the user to many other econometric methods with minimal programming; at the occasion the program has been made more modular and easier to maintain;

  • in the same vein, printing functions for univariate methods have been rationalized;

  • working of functions transforming variables into matrices (explon and explouniv), which are at the heart of almost all econometric functions, has been slightly modified for more flexibility: when a set of variables contains more than 1 constant variable, then a warning message is generated instead of stopping the function with an error;

  • functions stafore and dynfore could not deal with trends defined directly in the call of an estimation function: now they can;

  • calculation of forecast variance has been added to function varmaf

  • a few bugs have been corrected: working of functions performing temporal aggregation of time series (m2q, m2qa and q2a) has been restored for matrices of ts ('tsmat' objects); function pltseries could not deal anymore with time series and this has been restored; functions pltms_resid and pltms_yyhat did not adjust correctly the size of the sub-windows in the graphs presenting the residuals and the adjusted results from a multivariate Markov-switching and this has been corrected; the function performing the addition of 2 tsmats, %tsmat_m_tsmat did not work when the option for the names was 'reset' and this has been corrected.

    The user manual has been modified accordingly and slightly enhanced (correction of typos, of demos when they were no more working, ...).

    From the 1.5 to the 1.55 version

    (2013/04/04): Grocer version 1.55 has been adapted to the just released 5.4.1 Scilab version.

    (2013/03/16): Grocer version 1.55 has been posted on the web.

  • Function doornhans, that performs Doornik and Hansen normality test, has been extended to deal with multivariate residuals.
  • Ericsson and MacKinnon cointegration test was not properly implemented in function olsecm: now it is (Thanks to Bruno Quille for having pointed if to us).
  • Function varma did not accept more than 1 exogenous variable: now it accepts any number.
  • Disaggragation functions did not work properly when the high frequency data did not cover the whole low frequency period: now they do.
  • Function iv did not accept ts: now it does.

    (2012/11/10): the server that hosts Grocer's e-mails seems to be very tired and breaks regularly these times; so, if you send an e-mail at grocer.toolbox[at]free.fr, do not be surprised if the answer is unusually long; you can alternatively send your e-mails at grocer.toolbox[at]gmail.com (replace of course [at] with @).

    (2012/10/20): Grocer version 1.54 has been posted on the web.

  • Grocer functions have been adapted when needed to run properly with the 5.4.0 Scilab version.
  • Functions performing Insee's method for disaggregating time series, written by Emmanuel Michaux, have been added to the 1.54 version: the corresponding functions have circulated in French circles since a long time, in particular to mimic the building of the French quarterly national accounts; but, as examplified by a paper by Dorian Rouché and Michael Sicsic (www.tresor.economie.gouv.fr/File/375987), they can be used in other contexts. This method allows the adjustement of different low frequency models (AR(1) or Random Walk with or without constant) and then minimizes the sum of the difference of squared residuals, in order to minimize the fluctuations of the series stemming from the residuals. The corresponding function is etalcalinsee, with the corresponding demo etalcalinsee_d (see also the help file: help etalcalinsee).
  • Function hessian0 has been corrected to avoid a bug with function ms_quali.
  • An explicit error message when the # of switching variances is entered with a wrong value has been added to functions ms_mean, ms_reg and ms_var.
  • A special character in function IPS1 that made Grocer installation crash on some O/S has been replaced.
  • Aggregation functions for times series m2a, m2q and q2a have been changed in order to be able to deal with NA values (an NA value in a month does not prevent anymore the calculation of a quarterly aggregated series for quarters with non NA months).

    (2012/06/30): Grocer version 1.53 has been posted on the web.

  • Function msvar_irf, that aims at implementing impulse response functions for Markov-switching VARs, was not working properly: tests made by myself and Makram Khalil (thanks Makram) semm to indicate a proper working now.
  • Function automatic_signed, that performs automatic regression with sign constraints on the coefficients, was not working properly in some cases because of problems in the corresponding printing funtion: this is fixed.
  • Function sur, that performs seemingly unrelated regressions, contained a residual pause: this pause has been removed.
  • Function hessian0 has been extended to allow passing variable arguments of the input function.
  • In function brybos, the default algorithm when time series are neither monthly nor quarterly has been turned to 'bb' (for Bry-Boschan) instead of 'hp' (for Harding Pagan) because of the greater flexibility of this option.
  • Function kalman has been slightly modified: it now uses function hessian0 to calculate standard errors.

    (2011/02/18): Grocer version 1.52 has been posted on the web. This is a small update, whose main purpose is to make Grocer run on the forthcoming Scilab 5.4.0 version (already available as an alpha version). Emmanuel Michaux has also added a function waldchi performing a Wald test of linear constraints using its Chi-Square form (see function waldchi_d for a demo) and an option 'notstat', that prevents from calculating the Student statistics of the estimated coefficients, has been added to functions ms_mean, ms_reg and ms_var: this can be useful if the numerical hessian is not invertible (for instance in the case you have many coefficients) or in an experimental phase (moreover, you can calculate them ex-post with function MSVAR_stderr). Several bugs have also been corrected: function readxls2b has been made working on ts without comments; function varmaf has been made working when the user had not given a name to its variables in function arma (thanks to Jekaterina Borodina for having signalled and corrected this bug); function prtauto_signed which was lacking has been added; function gmmAndMon.sci, which used the wrong sign of the AR part when using an ARMA, now uses the good one; function BNG_ur did not work with more than 1 factor, now it works; option 'mindat' that was not correctly implemented in function explone is now correctly implemented; in function car2freq the good folder definition for the database basets.dat has been restaured; the difference of 2 tsmats (performed by function %tsmat_s_tsmat) is now working when the option 'reset' prevails.

    (2011/10/29): Grocer version 1.51 has been posted on the web. 2 new capabilities have been added:
    - a function msvar_irf that provide impulse response functions for Markov-switching VAR models: for a demo see function msvar_irf_d or help msvar_ird; since I have not been able to confront the results of this function with available results, I invite the user to check carefully the code and/or the plausibility of the results she may obtain.
    - a function contrib_logm2grq, built upon the model of function contrib_logq2gra, and which calculates contributions of exogenous variables to the quarterly growth rate of an endogenous variable when the econometric equation is estimated on the logarithm of the monthly observations of the variable (see the help from contrib_logq2gra for explanations).

    A few bugs have been corrected: in function adf, the order of the critical values, which was wrong, has been corrected; in function cadf, option 'noprint', which was not working, now works; help files for bfl and bfl1 were not correct (thanks to Cecile Ballini for having signaled all these bugs); function banerji works again, with the replacement of the removed Scilab function sort with gsort.

    (2011/09/28): Grocer version 1.5 has been posted on the web.

    The main novelties of the 1.5 with respect to the previous 1.46 version are the following:
    - you can now run ols with ARMA errors (function olsarma);
    - Andrews and Ploberger break point tests have been introduced, along with Bruce Hansen bootstrap method to evaluate their p-value (function AndPlob);
    - Basic quantile regression is now available in Grocer (function qreg);
    - those who do not want to use the bootstrap method with the Johansen cointegration method introduced with the 1.4 version can again use the asymtotic p-values (function johansen);
    - an imperfect, partial, but nevertheless substantial Gauss to Scilab translator (function gauss2sci);
    - Help files were not totally adapted to Scilab 5.3 family: they have been adapted and examples have been checked, corrected and improved at the occasion;
    - creation of a function automatic_signed that provides estimation à la pc-gets with sign constraints on the coefficients;
    - creation of a function quantile that calculates quantiles of colums of a vector;
    - adaptation of functions ecm and becm to allow the user to remove from the short term exogenous variables the part that can be integrated into the cointegration relationships;
    - taking into account Michaël Baudin's advice, non essential optimisation functions dfp_min, frpr_min and pow_min have been removed;
    - correction of several bugs: impexc2bd did not work when lines in the text files ended with a semicolumn; function mean did not work with matrices of ts; function uninstall_grocer did not work properly anymore; disagregation functions bfl, chowlin, denton, difonzo, fernandez and litterman did not work when bounds were given; since the extension in version 1.46 to more general cases, functions ms_mean, ms_reg and ms_var did not work anymore when a priori datation was given or when there were non switching variables; function fan_chart did not work when the number of bands was lower than 5; option bandwith='C' did not work in function ADF_individual; when LagOrders was given, function Levin_Lin did not work; function Pesaran did not work anymore starting from Scilab 5.3.0 versions: a small change has been made to restore the function; function oprobit1 did not work when not launched from function oprobit; function statfore did not work with ts when there was only a constant in the model; functions da2m, da2q and datelf2hf0 did not work anymore; many overloading tsmat functions did not work with 'reset' option.

    The manual has been updated and corrected: in particular, the names of many folders used in the manual examples, which were not correct anymore since Grocer adaptation to Atoms, Scilab module manager, have been restaured.

    From the 1.4 to the 1.46 version

    (2011/02/16): Grocer version 1.46 has been posted on the web. Following a request by Philipp Schuster, Markov swicthing regressions now allow exogenous variables to differ from an endogenous variable to another. Help files have been updated accordingly. Note that ascending compatibility has been respected as regards the high level functions ms_mean, ms_reg and ms_var, but not as regards many low level functions, such as ms_estimate. Sorry for any inconvenience, but this was imposed by the changing nature of the problems now dealt with the function ms_reg (if it poses any problem to you, do not hesitate to contact us by e-mail). The MS graphs have also been changed to show the smoothed residuals as well as yhat, instead of the filtered ones. The filtered ones are still available in the results tlist and can still be plotted (see the help for the functions pltms_resid and pltms_yyhat). Lastly, some examples in the help files were still making calls to the obsolete Scilab function mtlb_load (even if the matlab databases have been transformed into Scilab ones since a long time): the examples now make correct calls (thanks to Holger Nahrstaedt for having signalled this).

    (2010/11/01): Grocer version 1.45 has been posted on the web. Unless the 1.44 version, version 1.45 runs on the recently released 5.3.0 Scilab version. Several problems with help files on different versions have been corrected. Function nls was not working when bounds were set to empty through the command bounds(): this is corrected. The new function MSVAR_Setinit, used by functions ms_mean, ms_reg and ms_reg, did not work when there were common exogenous variables: it is corrected (thanks to Philipp Schuster for having signalled this bug). Division of a real by a ts was wrongly programmed: this is corrected. Graphs of regressions with discountinuous bounds were not working anymore: they work again.

    (2010/11/01): Grocer version 1.44 has been posted on the web. Functions ms_var, ms_mean and ms_reg have been enhanced along several directions: the initialization function MSVAR_SetInit has been modified (basically, the initialization rests on the residuals of the ols regression assuming no switching, against the values of the endogenous variables as before), which leads to more robust convergence; a bug in the calculation of the standard errors of the estimated coefficients has been corrected (thanks to Philipp Schuster whose tests and comparisons with its own results has lead to these improvements); lastly speed has been enhanced again, by a factor 2. There was a bug with function readxls2bd and functions da2m, da2q, date2fq, date2num_fq, def_dailyform, delts, lagts, num2date: these bugs are corrected (thanks to Zniber Housni for having pointed out these bugs; these bugs are the last, I hope, due to the modifications demanded by the Scilab team to integrate Grocer in Atoms, Scilab modules manager).

    (2010/07/03): Grocer version 1.43 has been posted on the web. Function ecm was not working in Scilab 5.2, because of a known Scilab bug regarding left matrix division: this problem has been solved with the replacement of the left matrix division with Grocer function invxpx. Function automatic was not working when there was only one specification test entered trough the option 'test=...": this is repaired. For Scilab versions starting from the 5.2, varma help chapter was filled with var help files instead of varma ones: this is fixed. Many functions have been somehow modified in order to improve their readability and speed: the improvement in speed is often marginal, but it is significant for functions ms_var, ms_mean and ms_reg.

    (2010/05/09): Grocer version 1.42 has been posted on the web. The installation device has been improved: as before, you can install Grocer by unzipping the instillation file into Scilab directory (the most user-friendly one, but with the drawback that it changes the scilab.star file, which is not recommended -rightly so!- by the Scilab team); but, you can now also do that by unzipping it into Scilab contrib folder, which creates a Grocer item in the toolboxes menu and you can install it in any library and using the Scilab standard procedures, that is running the builder.sce and loader.sce files.
    Function readxls2b has been improved in order to allow adding comments to ts, as with impexc2bd function. Function impexc2bd has been made slightly more robust: the function suppress useless quotes (which happens for instance when the csv file comes from OpenOffice); it replaces commas with dots and lastly the function signals when names begin with a figure (which is not allowed in Scilab). Function pltfac_kalm did not work with factors estimated on matrices: this bug is corrected (thanks to Ginters Buss to have signalled this bug). Scilab). Function pltfac_kalm did not work when there was more than one factor: this bug is corrected (thanks again to Ginters Buss to have signaled involved only one coefficient: this bug has been corrected.

    (2010/02/05): Because I have noticed a problem of compatibility for help files between the 5.2 Scilab version and the older 5.X versions, Grocer now comes under 3 distributions: a distribution for Scilab versions until the 4.1.2, a distribution for Scilab versions between 5.0 and 5.1.1 and a distribution for Scilab versions from 5.2.0 .

    (2010/01/31): Grocer version 1.41 has been posted on the web.
    - As the 1.4 one, this version works with Scilab 5.2.0, but, unless the previous one, the help takes advantages of the possibility offered by Scilab 5.2.0 to organize help files under a general chapter, Grocer in this case (thanks to Pierre Maréchal for having provided the stuff used to that end).
    - Run time of function fac_kalman has been greatly improved.
    - The adaptation of function and q2a to tsmat had lead to bugs with respect to ts: they are now corrected and a function m2a performing annual aggregation of monthly time series has been introduced at this occasion.
    - Because of change in Scilab function "who", function dblist had been modified and its working changed in a subtle way: the variables were presented in lexical order in the 1.4 version, which did not preserve ascending compatibility (thanks to Ginters Buss for having signaled this change). The old working has been stored in the 1.41 version.
    - The working of function varf had also been changed incidentally: the user was now obliged to enter the values of the constant variable over the simulation period: the old working has been restored (thanks again to Ginters Buss for having signaled this change).
    - Function ms_reg did not work when there was no switching exogenous variable: this bug has been corrected (thanks again to Ginters Buss for having signaled this bug).
    - Function ms_forecast did not work with results from a ms_var estimation. This bug has been corrected (thanks to Ginter Buss for the correction!).
    - Grocer has become Scilab Partner External Module (see http://www.scilab.org/pem/) and Grocer can now be download through the Scilab menu Atoms (from Scilab 5.2.0 only). Note that only the 1.4 version can at this date be download this way, but it will be soon the case for the 1.41 version.

    (2009/12/06): Grocer version 1.4 has been posted on the web. The main novelties of the 1.4 with respect to the 1.331 version are the following:
    - a new type of object has been introduced: matrices of time series that allow mass calculations with time series;
    - a field "comment" can now be added to time series, in particular when importing time series from an Excel or an ascii file;
    - most Johansen cointegration tests (weak exogeneity, test on the cointegration vectors or on the error correction coefficients,...) have been introduced and the Johansen cointegration method now allows any kind of exogenous variable, in the long run as in the short run;
    - many panel unit root tests have introduced;
    - Markov-switching methods applied to qualitative variables have been introduced;
    - a general function (dealing in particular with time series) allowing multiple graphs on the same window has been introduced;
    - a bug in the calculation of the gradient has been corrected in function fac_kalman (thanks to Laurent Ferrara for having signaled this bug);
    - a few bugs in functions ecm, becm, bvar and varma have been corrected (thanks to Ginters Buss for having signaled these bugs and corrected most of them!).

    From the 1.3 to the 1.331 version

    (2009/07/06): Grocer version 1.331 or Scilab versions starting from the 5.0.2 or for older Scilab versions has been posted on the web. This version includes an extension of function automatic that allows the Newey-West correction for standard errors. A bug in function pltseries (the option 'yaxis=...' did not work anymore) has been corrected (thanks to Ginters Buss for having signalled this bug).

    (2009/06/15): Grocer is to be presented at ScilabTec’09, the first Scilab Users’ Conference, in Supelec, on July the 1st, 2009.

    (2009/05/03): Grocer version 1.33 for Scilab versions starting from the 5.0.2 or for older Scilab versions has been posted on the web. This version includes a new function dynfore that performs the dynamic simulation of an equation estimated by ols, hwhite,etc. Function fac_kalman was not working in many cases since more than one factor had been allowed: it now works in those cases. The function fac_acp, fac_acp1, fac_acp2 and fac_acp_d (acp=analyse en composantes principales = principal component analysis in French) have been renamed to fac_pca, fac_pca1, fac_pca2 and fac_pca_d.

    (2009/02/15): Grocer version 1.32 for Scilab versions starting from the 5.0.2 or for older Scilab versions has been posted on the web. This version is the first one to work with the just released Scilab 5.1 version. Several bugs have also been corrected: for functions johansen, ecm, becm, garch, the option 'dropna', did it works, now they work (thanks to Issei Koduru for having signalled this bug); in function twosls, the Student statistic of endogenous rhs variables were incorrectly reported, now they are correct (thanks to Nicolas de Riccardis for having signalled this bug); function varmaf was broken, it now works again.

    Note that Grocer versions for the Scilab 5.0 and Scilab 5.1 versions and for older Scilab versions come with different zip files. This is because help files are handled differently according to Scilab version. Note also that help files for the 5.0 and 5.1 versions have been translated semi-automatically, so that some help files in the Scilab 5.0 version surely need to be corrected. Do not hesitate to signal us files that need so. Note lastly that the presentation of Grocer help with Scilab 5.0 and Scilab 5.1 versions can probably be improved: this is left for future release.

    (2008/10/18): Grocer version 1.31 for older Scilab versions has been posted on the web. The function fac_kalman has been extended to deal with any number of dynamic factors. A few functions dealing with ts have been added to allow checking if 2 conditions on ts are simultaneously statisfied (ts1&ts2 where ts1 or ts2 is itself the result of a test such as ts1==3 or ts2>5) or if at least one of the 2 conditions on ts is statisfied (ts1|ts2). A function fan_chart producing fan-charts à la Bank of England has been added. Function ms_var did not work when the size of the var was greater than 1; now it works (thanks to James Eustace for having signalled this bug). Function fac_acp did not minimize properly the information criterion when the number of factors is determined by the Bai and Ng criterion: now it is. Several help files have been corrected.

    (2008/04/06): Grocer version 1.306 has been posted on the web. The function studentize has been extended to deal with time series. The function var (lower case) has been changed to VAR (upper case) to avoid side effects with some Scilab functions (var is indeed a convenient name for a variable...). Bugs to functions ms_reg and ms_forecast have been corrected: functions ms_reg and ms_forecast did not work when there were no common exogenous variable (thanks to Thomas Raffinot for having signaled these bugs). The case when the starting conditions for the Kalman filter were not stationary was not dealt in function smooth_innov: now it is; and the convergence of the dynamic factor function fac_kalman has been improved by imposing the variances to be positive (thanks to Roumen Vesselinov for having provided the example that fostered these improvements).

    (2008/04/06): Grocer version 1.305 has been posted on the web. Minor improvments have been brought to the function automatic: the strategy followed and its parameters are not displayed with the final model; the arch test can now be added to the list of tests without going through the option 'newfunc='; Some minor errors to the help files relative to the functions automatic and bma_g have also been corrected.

    (2008/03/18): Grocer version 1.304 has been posted on the web. A bug had been introduced in function varma and mysteriously not detected; it is corrected (thanks to Cyril Voyant for having signaled us this bug). The function fac_kalman did not work for white noise residuals, nor for moving average (MA) factors (thanks to Laurent Ferrara for having signaled us this bug): this is corrected. The function prtts now checks properly if the frequencies of the printed time series are identical.

    (2008/02/02): Grocer version 1.303 has been posted on the web. A bug in function brybos has been corrected: when the bounds did not correspond to the time period of the series, the field 'phases' was not correctly created; now it is (thanks to Christelle Minodier for having signaled us this bug). Fields 'ans_vari_to_shockj' have been added to function irf to facilitate the recovery of irf results.

    (2008/02/02): Grocer version 1.302 has been posted on the web. A typo in function statfore has been corrected (thanks to Nicolas de Riccardis for having signaled us this bug).

    (2008/01/23): Grocer version 1.301 has been posted on the web. A bug introduced in function acf has been corrected (thanks to Juan-Pablo Ortega for having signaled us this bug).

    (2008/01/22): Grocer version 1.3 has been posted on the web. The main novelties of the 1.3 with respect to the 1.224 version are the following:
    - daily and weekly time series have been introduced and most econometric programs now can eliminate non-NA values from the estimation period
    - a much more efficient function for importing Excel files
    - a more efficient (3 to 4 times more rapid) varma function which now also provides the estimated residuals
    - calculation of descriptive statistics
    - small sample correction has been introduced in the White's adjusted heteroscedastic consistent Least-squares Regression (a correction due to Nicolas Sauter)
    - the Box-Pierce and Ljung-Box tests of autocorrelations and Li-McLeod tests of squared autocorrelations
    - the General Method of Moments has been introduced
    - Static and Dynamic factor estimation is now available
    - the cointegration test proposed by Ericsson and MacKinnon has been integrated
    - new econometric methods dealing with qualitative variables have been introduced: multivariate logit (translated and adapted from a Matlab program by Simon D. Woodcock) and ordered logit and probit with any number of categories
    - the manual and help files have been extended accordingly and corrected; some broken links to individual help files have been restored
    - A bug in Clark-West test for nested models has been corrected: the estimation of the HAC variance was always performed with a truncation window of 1 whatever the user asked for; now the user can impose her own soze for the truncation window (thanks to Hélène Erkel-Rousse for having signaled us this bug).
    - A bug in the garch function has been corrected: contrary to what was indicated, garch(p,q) models with p or q different than 1 could not be estimated; now they can (thanks to Juan-Pablo Ortega for having signaled us this bug).

    From the 1.2 to the 1.224 version

    (2007/11/12): Grocer version 1.224 has been posted on the web. The installation file did not work on Mac OS. This bug has been corrected by Stéphane Motellet (thanks Stephane).

    (2007/10/28): Grocer version 1.223 has been posted on the web. It is the adaptation of the 1.221 version that works with the just released 4.1.2 Scilab version. Function ecm has been corrected to allow a trend or nothing in the cointegrating relation (previously, as in James LeSage toolbox for matlab, there was automatically the constant and only the constant in the cointegrating regression).

    (2007/10/06): Grocer version 1.221 has been posted on the web. Several bugs have been corrected and some improvements incorporated: - the function varmaf that make forecasts form an arma estimation did not work anymore; now it works again (bug signalled by Eduardo Rodriguez).
    - minor improvements have been made to fuction automatic: now the paths followed by the models in stages 1 and 2 are displayed together with the display of the corresponding models.
    - display of funtion ers did not work properly; now it works.
    - a minor improvement to the accuracy of function statfore has been made.

    (2007/07/18): Grocer version 1.22 has been posted on the web. Several bugs have been corrected and some improvements incorporated:
    - when kpss was run after the bounds have been set to empty by running bounds(), then the program did not perform the test; now it does.
    - function rolreg did not work properly with vectors; now it works; moreover, the user can now impose the regression to withdraw variables whose coefficients have the wrong sign; the help file has been modified accordingly (all modifications made by Emmanuel Michaux).
    - fuction waldf did not work with models estimated by ols_cons; it now works.
    - fuction prtbanerji did not display the results in some cases; it is corrected.
    - fuction varma did not work anymore with exogenous variables; it works again.
    - the panel functions pbetween, pfixed and prandom worked with panel tlists, but not with matrices as indicated in the help files; now they work also with matrices (thanks to Michel Pouchain for having signaled the bug).
    - functions twosls and threesls did not work when the user did not enter the coefficients explicitely and when there were more than 10 coefficients; now they work also in that case (thanks to Meera Sharma for having signaled the bug).
    - function spectral did not work with vectors entered without quotes; now it works (thanks to Arun Tangirala for having signaled the bug).
    - the functions ms_reg and ms_var did not work when an a priori datation was given; now they work. Function ms_forecast did not work when exogenous variables over the forecasting period were entered as vectors; now it works (thanks to Tahoora Kh for having signaled this bug).
    - the 1.2 manual did not give credit to the true author of the matlab disagregation programs (Enrique M. Quilis) at the basis of grocer ones. This has been changed. Sorry, Enrique and thanks for your matlab programs!

    (2007/05/28): Grocer version 1.216 has been posted on the web. Ricardo Querall signalled us that the function adf did not work when the number of observations was too small. I have corrected this and taken the occasion to substitue the tables provided by Cheung and Lai (1995) to the previous ones, taken from James LeSage matlab toolbox (thanks Ricardo!). The only drawback is that the function adf does not work anymore with a time trend of order greater than 1 (if it is a problem, tell me: I can perform the necessary simulations). Eduardo Rodriguez has discovered a new bug in function varma: the function did not converge in some cases; this was due to an improper treatment of the case when some roots of AR part were inside the unit circle. This is corrected (thanks Eduardo!). Olivier Darné signalled us that the function ols did not work when the constant was the only exogenous variable. Although it is a rather uninteresting case, it had to be corrected: this is done, as well as in functions ols_cons and rolreg (thanks Olivier!).

    (2007/05/17): Grocer version 1.215 has been posted on the web. A small bug in function bma_g has been corrected (the option 'noprint' did not work, it now works; thanks to Florent Pochon for having signalled us this bug). A new option has been added to function expbd2exc that allows the user to choose the separator in the exported file. Note that the release of the new Scilab version 4.1.1 does not make any change to the functionning of grocer: any grocer version working with Scilab 4.1 version will work with Scilab 4.1.1 version.

    (2007/05/01): Grocer version 1.214 has been posted on the web. Various bugs discovered by Eduardo Rodriguez (thanks!) in varma estimation (various "global" variables were not propagated with the good name and made the function collapse) and in the function expbd2exc (the option 'transpose' did not function anymore) have been corrected.

    (2007/04/04): Grocer version 1.213 has been posted on the web. The option 'noprint' in function adf did not work in previous versions: it now works (thanks to Eduardo Rodriguez for having signalled me this bug). Function varma has also been slightly modified: it now allows the user to enter an option 'prtopt' that provides the values of the options used for the filering steps (more to come in the 1.3 version).

    (2007/02/10): Grocer version 1.212 has been posted on the web. The function impexc2bd has been modified to allow: dates not entered in chronological order (upon a request by Michel Pouchain); the user to enter her own definition for the dates name and the NA values. A bug in function automatic, linked to the "reliability" extension, has been corrected: when the final model was empty, the function could not calculate the reliability of any variable and crashed (thanks to Christelle Minodier for having signalled me this bug).

    (2007/01/14): Grocer version 1.211 has been posted on the web. The previous versions from the 1.202 to the 2.1 contained 2 versions of the function pltbrybos and the installation device unfortunately compiled the wrong version. This is repared. This version contains also a new funtion uninstall_grocer (and the corresponding help file) that allows the user to uninstall the current Grocer version (before installing a new one). I highly recommend, in particular for those who have download Grocer versions between the 1.202 to the 1.21, to install the 1.21 one, to run the function grocer_uninstall and then to install again Grocer. It will also be more secure to uninstall grocer 1.211 by this function when installing new Grocer versions.

    (2007/01/07): Grocer version 1.21 has been posted on the web. A function ms_forecast providing forecasts from a Markov-switching model has been incoporated (upon a request from Subhash Subramanian). The result tlist from the Bry-Boshan function has been completed by a field 'phases' that takes the values 0 in a recession and 1 in an expansion (upon a request from Slimi Souhir). The convention provided at the installation (peaks belong to exapnsions, troughs to recessions) can be changed by function define_recession. On-line help has been extended to incoporate these modifications. Functions allowing comparisons between ts and numbers have been created: ts<a (resp ts>a, ts<=a, ts>=a) with a a real number now provides a ts equal to 1 if the condition is true, 0 if false (an converselly for a<ts, a<=ts,...). The function prtts has been modified to avoid unfortunate side effects. The estimation of varma models has been modified and its speed somehow improved (note that it entails modifications to the function lffast: if there are grocer users that make use of the function lffast, they will perhaps be obliged to modify their programs; tell me if it is really a problem).

    (2006/12/23): Grocer version 1.203 posted the 20th of december exhibited some strange behaviour. Although I have not fully understood the source of this problem, I have recreated the installation file and it seems to work properly again.

    (2006/12/20): Grocer version 1.203 has been posted on the web. Bugs to function pltspectral (thanks to Jamel Gaftaoui for having signalled this bug) and impexc2bd (thanks to Othman Bouabdalah for having signalled this bug) have been corrected.

    (2006/12/17): Grocer version 1.202 has been posted on the web. Bugs to function pltbrybos and banerji have been corrected. Small improvements to function impexc2bd have been made (changes any numerical separator ',' 'into '.'; accepts blanks in a the 'dates' name). Function freqts_c used by function expbd2exc, that has been omitted in the 1.201 version, has been added. Note that this version works with any Scilab version from the 3.0 to the -just released- 4.1 one.

    (2006/11/19): Scilab.star file was corrupted. It is now repared. All my excuses to all grocer users for all the problems of last week: I will be careful to avoid them to happen again on grocer future versions.

    (2006/11/15): Scilab help for previous 1.201 grocer version was no more available. It is now available again.

    (2006/11/12): Version 1.201 has been posted on the web. This is the first version to work with the -just released- Scilab 4.1 Release candidate version. The function varmaf and the corrresponding help has also been modified to better deal forecasts with time series. A function hprev2vec used as a subroutine in forecasting functions has been created. The help for the function robust has been corrected (thanks to Emmanuel le Meur for having signaled this error).

    From the 1.1 to the 1.109 version

    (2006/09/03): Version 1.109 has been posted on the web. A bug has been corrected in function prtjohan (the function could not print stored tlist results: now it can; thanks to Carlos C. Aranda for having signaled it to me and Emmanuel for having corrected it!). Tables of the cusum-squared test provided by Eddgerton and Wells have been incorporated to the cusumb and cusumf functions upon a suggestion by Pascal Grandeau (thanks for having sent me the article I did not know!). Constrained ordinary least squares (functions ols_cons2 and ols_cons) have been added upon a request by Michel Pouchain. This should be the last version before the enhanced 1.2 version (adapatation to the new graphic mode, Markov-switching models, Bayesian Model Averaging method, improvements of various existing functions and a few other new ones...) due towards the end of october.

    (2006/07/03): Version 1.108 has been posted on the web. Bugs have been corrected in function gamm_rnd (an improper translation from Matlab led to a wrong distibution; thanks to Mario Maggi for having indicated it to me), olsar1 (an error was generated because the numbers of calls to olsc0 was limited to 10; thanks to Michel Pouchain for having indicated it to me) and auto_stage0 (when all variables where insigificant at the loose significance level used at the second step of automatic to eliminate globally insiginficant variables, the programm tried to estimate an empty model; thanks to Pierre-Emmanuel Ferraton for having indicated it to me).

    (2006/04/12): Some users working on Windows98 have signalled me that grocer insatllation generated the following error message:
    !-- error 49
    incorrect file or format
    at line 68 of function genlib_grocer called by :
    genlib_grocer()

    I have therefore developped an installation device that works on Windows98 and should work in other cases when the standard installation device does not work. It is available here

    (2006/03/15): Version 1.107 has been posted on the web. Corrections have been brought to the var and prtvar functions (again, correction of bugs due to changes from the 1.0 to the 1.1 version). Contrary to what the help file indicated, the function ts2vec did not accept bounds entered as a row; now it does (for progammers!). Note that this version should work with Scilab versions from the 3.0 til the 4.0.

    (2006/01/17): Version 1.105 has been posted on the web. Corrections have been brought to the panel data functions (thanks to Michel Pouchain for having signalled me the bugs) and the johansen function (that did not work anymore with ts, due to changes from the 1.0 to the 1.1 version). Emmanuel Michaux has added to the function var and var1 the option to withdraw the constant, which remains by default automatically included.

    (2005/12/30): Version 1.104 has been posted on the web. This version works on the just released Scilab version 4.0-RC1. I however recommend using the older Scilab versions 3.0 or 3.1.1, because the 4.0-RC1 is only a "Release Candidate" one and remains therefore to be extensively tested. A few changes have been also brought to the 1.102 version: display of automatic (variables are now displayed according to the entry order given by the user) and varma (when the VARMA is a true VARMA and not an ARMA, the display is clearer) has been improved; the variance matrix of residuals in a varma estimation, which waw rightly calculated but wrongly reported, is now correctly so; the output result from varma now contains in the fields 'AR', 'MA', 'ARS', 'MAS', 'V' and 'G' the estimated values instead of the starting values; in all the programs that used hypermatrices, I have replaced them by simple matrices (this is a little bit less readable, but more efficient); lastly Emmnanuel Michaux has corrected his spectral functions (the calculation of the cospectrum is now only performed when there are more than one series; all the results involving multiple series -such as cospectrum, cohesion,...- are now plotted once there are at least 2 series and not 3 as before).

    (2005/12/11): Version 1.102 has been posted on the web. This version corrects several bugs on varma estimation discovered by various users (Emmanuel Michaux, Patrick Lee and Reza Solgi: thanks to all): in particular the programm did not work for seasonal ARMA, nor with time series (due to an improper adpation to the changes in the treatment of data from the 1.0 to the 1.1 version); now it works.

    (2005/11/25): Version 1.101 has been posted on the web. The changes from the 1.1 version are very limited: a bug (signaled to me by Pierre-Emmnanuel Ferraton: thanks) in function bkwols has been corrected (the function did not work properly in the very peculiar case when there was no other exogenous variables than the constant); in functions litterman, fernandez and chowlin, the output is now a ts over the same period as the indicators when these are temselves ts (upon a suggestion by Emmanuel Michaux)

    (2005/11/09): Version 1.1 has been posted on the web. The main changes from the 1.041 version are the following:
    - new business cycle procedures, due to Emmanuel Michaux: spectral analysis, the Bry-Boschan-Harding-Pagan procedure for the datation of turning points, Banerji's test of lead at turning points;
    - addition of basic panel data methods, mostly translated and adapted from matlab programs written by Carlos Alberto Castro: fixed and random effect, between estimation, Haussman test...
    - addition of several procedures, mostly translated and adapted from matlab programs written by Enrique M. Quilis, to disaggreagate low frequency timeseries to high frequency ones: Chow-Lin, Litterman,... methods;
    - several improvements of the pc-gets like program automatic: big increase in speed (up to tenfold) due to a major improvement in the heteroskedascity test; implementation of the so-called 'liberal' and 'conservative' strategies; addition of a bottom-up estimation and estimation of groups of variables with a given significance...
    - a slight modification to bkwols that allows to detect exact to extreme colinearity and sends a warning message to the user
    - a new feature that allows the user to define her own time series frequencies;
    - the manual has been updated accordingly.

    From the 1.0 to the 1.041 version

    (2005/20/08) I have been told that the installation of the 1.04 version did not always work properly, although it worked perfectly in the environments that I have tested. So I have posted Grocer 1.041 on the web with a slighty different installation mode, which I hope will not generate the same problems as the 1.04.

    (2005/15/08): Grocer 1.04 has been posted on the web. The main difference with the 1.036 version is that it now runs under Scilab-3.1.1 (it continues to run under Scilab-3.0). This has constrained to modified numerous demo funtions (lad_d, tobit_d, sur_d, tvp_d1, all varma demo functions,...), theilbv, drawy, pltseries0. A typo in function explovars has been corrected and varcov0 has been corrected (varcov0 was created for centered variables, it can now deal with non centered variables).

    (2005/06/06): Grocer 1.036 has been posted on the web. Corrections have been made to functions dfbeta (thanks to Emmanuel Michaux for the correction), probit (thanks to Mauro Barrera for the correction), tobit (all three functions worked only with some type of data) and ols1a (thanks to Pierre-Emmanuel Ferraton for having signaled me the bug: in some estimations performed by automatic yhat was not saved; now it is).

    (2005/05/04): Grocer 1.033 has been posted on the web. It replaces the 1.032 version that was deficient on 2 aspects: the file bkwols.sci was mysteriously lacking and the file explots.sci was not the good one (I have adapted my procedures to avoid such a mistake to be repeated...).

    (2005/04/20): Grocer 1.032 has been posted on the web. The differences with the 1.03 version are the following:

  • New corrections on the installation of Grocer on linux systems.
  • Correction of a small bug in function nls.
  • Another extension of function impexc2bd: the function now accepts incomplete dates (only one date is now sufficient: the function fills the blanks

    • (2005/03/21): Grocer 1.03 has been posted on the web. The differences with the 1.02 version are the following:

  • The installation of Grocer on linux systems did not work properly in particular because linux does not read blanks at the beginning of the lines in .sci files functions. I hope it now works (thanks to Martin Panggabean for having signaled me this bug and helped fix it). Functions genlib_grocer, cffilter, prtvar, kalman, tvp, maxlik, pow_min, dfp_min and frprmin have been modified accordingly.
  • Instrumental variables estimations have been robustified with respect to ts. Functions explots and explovars have been created to that end and function explouniv modified accordingly.
  • function automatic has been improved on various details: in particular, you can now impose the presence in the regression of the constant alone; functions auto_stage0 and auto_stage0 have been modified to that end.
  • function impexc2bd has been slightly improved: non numerical values are now replaced by NA's and a test has been added to check if the separator is the good one
  • functions nwest and nwest1 (provided by Emmanuel Michaux), that compute Newey-West's adjusted heteroskedastic and autocorrelation consistent standard errors, have been added. Function prtres has been modifed accordingly.
  • a small bug in function diebmar has been corrected (thanks to Emmanuel Michaux): the name grocer_y has been substituted to y at line 94.

    • (2005/01/30): Grocer 1.02 has been posted on the web. The differences with the 1.01 version are the following:

  • function automatic now allows the user to impose the presence of a set of variables in the final regression (it can be useful for instance to ensure the presence of a constant term in a regression); it has been submitted to intensive testing and residuals bugs have been found (there was notably a typo in auto_satge1 that prevented the program to work when no variable could be found significant: it is repared). Function has been adpated accordingly.
  • function ols0 now uses the function lsq that exists in Scilab-3.0 version and did not exist in older versions. The writing (but not the functionnalities) of some functions (robust, iv1, hwhite, dfbeta and recresid) have been adjusted accordingly.
  • functions olsc0, olsc1 and olsar1_1, less sophisticated and more efficient variants of functions olsc and olsar1, have been created and the on-line help for these functions as well as for functions olsc, adf and prtdiebmar (which were lacking) have been added.
  • numerous functions (hpfilter, cffilter, bkfilter, contrib,...) have been adapted to be able to treat bounds set to [] by the command bounds().

    • (2005/01/09): Grocer 1.01 has been posted on the web. The differences with the 1.001 version are the following:

  • correction of bugs in functions lfmodini (wrong treatment of explosive roots), automatic (when there was no significant variable, the package tried nevertheless to estimate an empty model; now it recognises that the model is empty- thanks to Emmnauel Michaux; in stage 2, when a model was equal to the union model, the programm did not notice and tried to perform a Wald test, which was not posible; this is corrected -thanks to Pierre-Emmanuel Ferraton)
  • function explox has been modified and its functionnality somehow changed (sorry for the ascending compatibility...); the treatment of bounds is now the following: if the user enter bounds in the function (3rd argument), they will apply; if no bounds are entered or if the bounds [] are entered, then: a) if there exist bounds in the environment, they will apply; b) if not, the bounds will be determined by the function; function explouniv has been modified accordingly; calls to the econometric functions are unchanged
  • addition of 3 new functions:
    - olspec providing both the ordinary least squares and 5 default specification tests or ones chosen by the user; a function test_spec, that calculates these 5 specification tests has been created and the function prtuniv modified accordingly
    - diebmar, provided by Emmnanuel Michaux, providing the Diebold-Mariano test of forecasting performance equality
    - varmaf, that provides forecasting with an arma; this function relies on a function foremod translated and adapted from matlab package e4
    • ; the functions varma, prtvarma and prtvarf has been adapted accordingly
  • functions waldf and waldf0 has been slightly modifed: it can now test an empty model against another one (an empty matrix must then be given as first argument)
  • the following functions have been modifed to avoid useless warnings: var1, irf0, irf_mc1, sur, kalman, tvp, smoothing, legends and pltseries0
  • speed of the function automatic has been improved by around 10%: functions auto_stage0, auto_stage1, bpagan0, hetero_sq0, arlm0 have been modified to that end
  • correction of various typos or small mistakes in the on-line help

    • (2004/11/11): Grocer 1.001 has been posted on the web. The differences with the 1.0 version are the following:

  • bugs have been corrected in functions bpagan (degrees of freedom were not correctly calculated), adf (the test without a constant term did not work; thanks to Abdenor Brahami for having signaled me this bug) and varma (it did not work when an exogenous variable was added- thanks to Emmanuel Michaux for having signaled me this bug);
  • minor changes (without any consequence with respect to their functionning) have been made to functions bino_rnd and %ts_p.