Some users might be unfamiliar with updates made in the previous releases. SAS/ETS 13.2 introduced the SASEQUAN interface engine. Following are highlights of the other enhancements in SAS/ETS 13.2:
The COUNTREG procedure was enhanced with the following Bayesian estimation features: the multithreaded Metropolis sampling algorithm, convergence diagnostic tools, the automated reiterative MCMC algorithm (which can search for the stationary region of the posterior distribution and determine the length of posterior sample), and a weighted scheme that can recombine the samples from multiple MCMC chains. The PERFORMANCE statement and item store support for BY groups were also added to the COUNTREG procedure.
Support for panel data analysis was added to the HPCOUNTREG procedure.
The HPSEVERITY procedure was enhanced to support the CLASS statement, classification and interaction effects in the SCALEMODEL statement, OUTSTORE= and INSTORE= options to store and read models that contain richer scale regression effects, PLOTS= and OUTCDF= options for single-machine mode, specification of probability of observability in the LOSS statement, and a richer set of options in the NLOPTIONS statement.
The following enhancments were added to the PANEL procedure: first-differenced methods for one-way and two-way models, cross-sectional dependence tests, Lagrange multiplier (LM) tests for cross-sectional and time effects, and tests for serial correlation and cross-sectional effects.
Enhanced Bayesian estimation features were added to the QLIM procedure. These features include an automated reiterative MCMC algorithm (which can search for the stationary region of the posterior distribution and determine the length of posterior samples) and a weighted scheme that can recombine the samples from multiple MCMC chains.
The SASEFRED engine was enhanced by adding error checking to avoid situations in which the UNITS= option, the OUTPUT= option, and the VINTAGES= option are incompatible.
The SEVERITY procedure was enhanced to support the CLASS statement, classification and interaction effects in the SCALEMODEL statement, OUTSTORE= and INSTORE= options to store and read models that contain richer scale regression effects, and an INITSAMPLE= option to limit the number of observations used for initialization that requires EDF computations.
The VARMAX procedure was enhanced by adding p-values for the Johansen cointegration rank test and a multistep forecast for multivariate GARCH models.
The X12 procedure was enhanced by adding the ALPHA= option to the FORECAST statement to specify the size of forecast confidence limits; adding the ALMOST=, ALPHA=, CVMETHOD=, LSRUN=, METHOD=, and TCRATE= options in the OUTLIER statement to enhance outlier detection; and by enabling you to specify seasonal filters in the SEASONALMA= option in the X11 statement for each seasonal period.