Release Notes for S+FinMetrics 2.0 (April 2005) These Release Notes are current as of March 1, 2005. ************************************************************************ Contents of these Release Notes ************************************************************************ * SUPPORTED PLATFORMS AND SYSTEM REQUIREMENTS * Windows-specific requirements * UNIX/Linux-specific requirements * INSTALLATION NOTE * WHAT'S NEW * FIXED BUGS * RUNNING S+FINMETRICS 2.0 IN S-PLUS * KNOWN ISSUES * CONTACT INFORMATION FOR FEEDBACK ************************************************************************ SUPPORTED PLATFORMS AND SYSTEM REQUIREMENTS ************************************************************************ This released version of S+FinMetrics 2.0 requires S-PLUS 7.0 Professional Developer or S-PLUS 7.0 Enterprise Developer on all platforms. ************************************************************************ WINDOWS-SPECIFIC REQUIREMENTS ************************************************************************ S+FinMetrics 2.0 for Windows is supported on the following platforms: Windows 2000 Windows XP Home Edition Windows XP Professional Edition Windows 2003 Server running on Intel platforms S+FinMetrics 2.0 is NOT SUPPORTED under WIN32S (that is, Windows 3.1x), nor is it supported under Windows NT 3.51 or Windows 95, 98 or ME. The minimum recommended system configuration is the same as S-PLUS 7.x for Windows: Pentium III with 512MB of RAM. ************************************************************************ UNIX/LINUX-SPECIFIC REQUIREMENTS ************************************************************************ S+FinMetrics 2.0 for UNIX/Linux is supported on the following platforms and operating systems. The minimum recommended disk space for installing and running S-PLUS is also included: Platform Operating System ---------------------------------------------------- Sun SPARC Solaris 2.8, 2.9 on SPARC processors Intel/AMD x86 Red Hat Enterprise Linux WS 3.0 Note that S-PLUS no longer supports HP Alpha running Tru64, or UNIX 4.0F or 5.1A. Please also note that previous versions of the listed operating systems may function with S-PLUS, but they are not supported. ************************************************************************ INSTALLATION NOTE ************************************************************************ Install S+FinMetrics in a new directory. If you have an existing installation of S+FinMetrics 1.0, please install S+FinMetrics 2.0 in a new directory, for example: splus70/module/finmetrics ************************************************************************ WHAT'S NEW ************************************************************************ The following is a list of features new to this product release: 1. Enhancements to existing functions. * SsfFit: Return the numerical Hessian matrix and covariance matrix of estimates based on Hessian matrix. * SsfSim: allows time varying mOmega and more efficient use of KalmanIni(). * more state space model method functions (print, summary, coef, vcov methods for SsfFit etc.) * coint: Two more arguments are added to coint(): H and b. These allow for the linear restrictions of the type beta = H * psi or beta = (b, psi). See Johansen (1995). * print.coint: An optional argument "restrictions" is added to print.coint() to print the LR test of the above restrictions. * VECM: Allows user-specified cointegrating vectors. * VECM.form: Obtain the VECM form from a multivariate time series. * perpMat: Computes the "perp" matrix, i.e., orthogonal complement. * VAR: allowing exogenous variables in the RHS of the equation * Additional argument "pad" is added to tslag() to allow non-NA padding. * an updated Shapiro-Wilk test for normality (Royston 95) added to normalTest.default * Function enhancements in EVIS and EVANESCE libraries: - added options for users to model and plot only one tail (gpd.tail, gpd, plot.gpd, tailplot etc.) - fit.copula: covariance matrix of parameter estimates based on numerical Hessian matrix - IC.copulaFit: the IC method function for copulaFit objects - compare.copulaFit: gives all the IC of a list of copulaFit objects 2. New functions have been added: * autocorTest.lm: LM test for autocorrelation of regression residuals. * autocorTest.default: Allows LM test for OLS objects directly. * print.autocorTest: print method expanded for LM autocorrelation test. * varRatioTest: Variance ratio test for both homoskedastic and heteroskedastic errors. * print.varRatioTest: Print the variance ratio profile. * plot.varRatioTest: Plot the variance ratio profile. * unitroot: Support method="mpp" for modified Phillips-Perron tests, method="dfgls" for Dickey-Fuller test with GLS detrending, and method="ers" for the ERS point optimal test. In addition, if lags is not specified in the call, automatic lag selection which minimizes MIC as in Ng and Perron (2001) is then used. For the latter reason, three new arguments are added: min.lags, max.lags, penalty. When penalty=0, mic is used; when penalty=1, bic is used. * unitroot.k: Select the optimal lag length. * detrend: Detrend a time series using either GLS or OLS detrending. * unitroot.mpp: Compute modified Phillips-Perron test. This is called by unitroot() when method="mpp". * unitroot.dfgls: Compute Dickey-Fuller test with GLS detrending. This is called by unitroot() when method="dfgls". * unitroot.ers: Compute ERS test. This is called by unitroot() when method="ers". * print.unitroot: Works with MPP test and ERS test. * summary.unitroot: Works with MPP test and ERS test. * SETAR: Self-exciting threshold AR models. * TAR: Threshold AR models (with automatic search for optimal threshold variable and threshold value) * STAR: Logistic Smooth Transition AR models * MSAR: Markov Switching Autoregressive models * print, summary, plot, residuals, and IC methods of SETAR, TAR, STAR, and MSAR are added. * nonlinearTest: Test for threshold nonlinearity (BDS test, Tsay test, Hansen's sup-LR test, LM test against STAR nonlinearity) * Markov Switching State Space models * Simulation and estimation of Affine Term Structure Models (ATSM) using State Space Models * Generalized Methods of Moments (GMM) * Semi-Nonparametric Estimation (SNP) * Efficient Methods of Moments (EMM) * Simulated solution to Stochastic Differential Equations (SDE) ************************************************************************ FIXED BUGS ************************************************************************ The following are bugs noted in S+FinMetrics 1.0 and fixed in this release: * Shapiro-Wilks normalTest() cannot handle sample size larger than 2000. * Shapiro-Wilks normality test fixed in garch.stat() which corrects the wrong normality test p-value printout in print.summary.mgarch(). * qq-plot option in plot.garch broken for ged distribution. * Problems with KalmanSmo() and SsfMomentEst(). * term.struct with plot=T results in error. * tail.index example from reference manual fails. * pdl gives incorrect results. * incorrect p-value output in mgarch function (summary.mgarch) * incorrect R square calculations in mfactor * unitroot adf n test statistic corrected for the case of extended lags * interpNA fixed for unevenly spaced missing data * other fixes to enhance the code stability (for example, NLSUR etc.) * documentation (help file) bugs. ************************************************************************ RUNNING S+FINMETRICS IN S-PLUS ************************************************************************ After installation, to run the S+FinMetrics module from the Command line, enter: > module(finmetrics) In S-PLUS for Windows, you can also load S+FinMetrics by choosing File > Load Module > finmetrics. ************************************************************************ KNOWN ISSUES ************************************************************************ There are no known issues at this time. ************************************************************************ CONTACT INFORMATION FOR FEEDBACK ************************************************************************ Please feel free to contact us with questions about this release. Send all questions and general comments to bugs@insightful.com We are very interested in receiving your comments and suggestions for improving S+FinMetrics 2.0.