Below is a summary of the available Bayesian methods and capabilities and the procedures in which they appear:
| Product | Beginning Release | Procedure | Bayesian Methodology |
| SAS/STAT® | SAS® 8 or earlier | DISCRIM | Computes posterior probabilities of group membership and posterior probability error rate estimates. Specifies prior probabilities of group membership in the PRIORS statement. |
| GLM | The WALLER option provides multiple comparisons of means, minimizing Bayes risk under additive loss. | ||
| LOGISTIC | The CTABLE option estimates false positive and false negative rates as posterior probabilities using Bayes theorem. The PEVENT= option provides prior probabilities. | ||
| MIXED | The PRIOR statement specifies prior distribution for variance components and estimates their marginal posterior density. | ||
| NLMIXED | The PREDICT statement and the OUT= option in the RANDOM statement provide empirical Bayes estimates of random effects. | ||
| TPSPLINE | The OUTPUT statement can request Bayesian confidence intervals for smoothing spline estimates. | ||
| GAM | The OUTPUT statement can request Bayesian confidence intervals for smoothing spline estimates. | ||
| MI | Can perform an approximate Bayesian bootstrap imputation of missing values. Prior information for Bayesian estimation of means and covariances can be supplied by using the PRIOR= option in the MCMC statement. | ||
| SAS 9.2 | GENMOD LIFEREG PHREG | The BAYES statement requests a Bayesian analysis of the regression model using Gibbs sampling. The model parameters are treated as random variables, prior distributions can be specified, and inferences about the parameters are based on the posterior distribution of the parameters given the data. | |
| MCMC | MCMC is a general purpose Markov chain Monte Carlo simulation procedure that fits Bayesian models given arbitrary prior distributions for the model parameters and a likelihood function for the data. MCMC obtains samples from the posterior distributions, produces summary and diagnostic statistics, and saves the posterior samples in an output data set. | ||
| SAS/STAT 9.22 in SAS 9.2 TS2M3 | GENMOD | The Gamerman and independent Metropolis algorithms are new sampling methods available in the SAMPLING= option of the BAYES statement. | |
| PHREG | The Zellner g-prior is available for the regression coefficients. The random walk Metropolis (RWM) algorithm can now be used to sample an entire parameter vector from the posterior distribution. | ||
| SAS 9.3 TS1M0 | FMM | Can provide Bayesian analysis of finite mixture models using either a conjugate Gibbs sampler or the Metropolis-Hastings sampler. | |
| SAS/STAT 12.1 in SAS 9.3 TS1M2 | MCMC | Models missing values by default. RANDOM statement supports multilevel hierarchy to an arbitrary depth. Uses faster and more efficient sampling algorithms. | |
| PHREG | Bayesian frailty models | ||
| GENMOD | The Gamerman algorithm is now the default sampling mechanism except in cases for the normal distribution with a conjugate prior. | ||
| SAS/STAT 13.1 in SAS 9.4 TS1M1 | BCHOICE | Performs Bayesian analysis for discrete choice models. Multinomial logit, multinomial probit, and nested logit models are available. | |
| HPFMM | Performs Bayesian analysis via a conjugate Gibbs sampler if the model belongs to a small class of mixture models for which a conjugate sampler is available. | ||
| SAS/STAT 14.1 in SAS 9.4 TS1M3 | MCMC | New sampling algorithms are added for continuous parameters: the Hamiltonian Monte Carlo and the No-U-Turn Sampler. | |
| SAS/STAT 15.1 in SAS 9.4 TS1M6 | BGLIMM | Provides full Bayesian inference for generalized linear mixed models (GLMMs). | |
| SAS/QC® | SAS 8 or earlier | ADX application | Can perform Bayesian selection of active effects in saturated and nearly saturated designs. Prior probabilities of active effects can be specified. |
| BAYESACT function | Computes posterior probabilities that observations in a sample are contaminated with a larger variance than other observations. Computes the posterior probability that the entire sample is uncontaminated. | ||
| OPTEX | Can search for a Bayesian optimal design. Specifies prior precision values in the PRIORS= option in the MODEL statement. | ||
| SAS/ETS® | SAS 8 | VARMAX | Can fit the Bayesian Vector AutoRegressive model and the Bayesian Vector Error Correction model. Specifies prior information by using the PRIOR= option in the MODEL statement. |
| SAS/ETS 12.1 in SAS 9.3 TS1M2 | QLIM | The BAYES statement allows Bayesian estimation of most of the univariate models available in the QLIM procedure. | |
| SAS/ETS 12.3 in SAS 9.4 TS1M0 | HPQLIM | The BAYES statement allows Bayesian estimation of most of the univariate models available in the HPQLIM procedure. | |
| SAS/ETS 13.2 in SAS 9.4 TS1M2 | COUNTREG | The BAYES statement requests a Bayesian analysis of the regression model using Metropolis sampling. The PRIOR statement specifies the prior distribution of the model parameters. | |
| SAS/IML® | SAS 8 or earlier | TSBAYSEA function | Performs Bayesian seasonal adjustment modeling. |
| SAS/Genetics™ | SAS 9.1 TS1M3 | HAPLOTYPE | Haplotype frequency estimation. |
| † Book-Based Macros | Multiple Comparisons and Multiple Tests Using the SAS System | %BayesIntervals | Computes Bayesian simultaneous confidence intervals. |
| %BayesTests | A SAS/IML macro that computes Bayesian posterior probabilities for a set of free-combination tests. |
† See the associated book for information about the use and output from these macros.
For more information in general about Bayesian analysis, see the SAS Users YouTube video.