Featured news from SAS.

 

 ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ 
View as a web page
 
SAS The Power to Know logo
 
SAS Statistics and Operations Research News

Fall is nearly upon us here in Cary, but not according to the thermometer. I hope it is somewhat cooler where you are.

We had a fantastic time in Denver at JSM. Denver is a wonderful place for a convention, and I hope many of you had a chance to attend. The SAS booth on the Expo floor saw quite a bit of activity. Causal analysis is a hot topic. The new CAUSALMED, CAUSALTRT, and CAUSALGRAPH procedures garnered more than their fair share of questions, as did our latest foray into all things Bayesian—PROC BGLIMM. Randy Tobias, R&D Director for SAS/STAT®, tells me that there were dozens of sessions devoted to the topic of causal analysis this year. Most, if not all, of these sessions were packed—standing room only! SAS was right in the mix of things with popular Continuing Education courses and Computer Technology Workshops in both of these exciting areas of statistics.

Fall brings with it our regional SAS® users group meetings. If you are not familiar with the regionals, you can find out more information at SAS Users Groups. This page also highlights local users groups in the United States. Click the box for your region of the country to see if there is a local users group in your area. I will be traveling to Phoenix for the Valley of the Sun users group meeting and to Denver for the Denver users group meeting at the end of October. If you are not in the US, fear not! Check out Meetup or LinkedIn to find a local group in your city or country.

Phil Gibbs

Manager, Advanced Analytics Technical Support

 

Technical Papers

 

Introducing the BGLIMM Procedure for Bayesian Generalized Linear Mixed Models

PROC BGLIMM is a new, high-performance, sampling-based procedure that provides full Bayesian inference for generalized linear mixed models, new in SAS/STAT 15.1. This paper provides a brief overview of GLMMs; introduces important features, statements, and options in PROC BGLIMM; and covers high-level simulation and algorithm details of the procedure. It also presents three examples, from simple to complex, to demonstrate how to use PROC BGLIMM.

 

Unleashing SAS® Visual Data Mining and Machine Learning Models

SAS Visual Data Mining and Machine Learning can be used together with other SAS products to build and compare various predictive models. First, you use SAS Visual Data Mining and Machine Learning to create several models, and you choose one of them as your champion model. You can publish all these models to different destination types, such as Hadoop, Teradata, SAS® Cloud Analytic Services (CAS), and SAS® Micro Analytic Service. You use SAS® Embedded Process to score the data against these published models where the data reside. You can also register the models in SAS® Model Manager and compare them against other models to select a final champion model. Then you can test these models to validate them for scoring. If you notice a degradation in the model, you can retrain it. Retraining the model triggers a run of all the pipelines in the associated SAS Visual Data Mining and Machine Learning project, and the recalculated project champion is automatically registered in SAS Model Manager. In addition, you can score streaming data by using SAS® Event Stream Processing on the models that are registered in SAS Model Manager. SAS Visual Data Mining and Machine Learning also provides a scoring API that enables you to score models directly in Model Studio by using RESTful interfaces. This paper shows how you can unleash the full power of your models by taking advantage of the model processing capabilities in all these SAS products.

 

Technical Highlights

 
Video icon

Build Fully Connected Neural Networks with SAS® Studio Tasks and SAS® Viya®

Brian Gaines, a developer in the Advanced Analytics Division in SAS R&D, shows how you can use the tasks in SAS Studio to quickly develop code for estimating a fully connected neural network with the power of SAS Viya.

SAS® Global Forum 2020

Join us in our nation’s capital March 29 – April 1, 2020, at the Walter E. Washington Convention Center. We’ve got spectacular offerings and experiences designed with you in mind. Registration opens in October! More details coming soon.

Blue blog icon

The DO Loop

Distinguished Research Statistician Developer Rick Wicklin shows how to create a discrete heat map with PROC SGPLOT, offers tips about anchor points and rotated text in PROC SGPLOT, and demonstrates how to conditionally append observations to a SAS data set. He also presents the essential guide to binning in SAS as well as how to compute and visualize the similarities among recipes using cosine similarity.

 

Video icon

Partial Proportional Odds Modeling with the LOGISTIC Procedure

Bob Derr demonstrates what to do when PROC LOGISTIC fits the default proportional odds model to multilevel response data but the proportional odds assumption is rejected. He shows how to use the UNEQUALSLOPES option, introduced in SAS/STAT 12.1, and the EQUALSLOPES option, introduced in SAS/STAT 13.2, to fit a partial proportional odds model to these data. He also discusses graphical and selection methods to determine which covariates have proportional odds and shows how to use likelihood-ratio tests to test for goodness of fit of the models.

 

Tech Support Points Out

 
Tech Support Points Out icon

Nonconvergence in Log-Linked Poisson and Negative Binomial Models

Convergence failure is common in binary-response models, such as logistic and probit models, because of separation—the ability of the model to separate the event from the nonevent responses. Although this seems desirable, it results in some of the model parameters being infinite, which prevents convergence. This issue in binary-response models is further discussed, along with solutions, in Usage Note 22599 and in the PROC LOGISTIC documentation. A similar convergence problem can occur in log-linked count models, such as Poisson and negative binomial models. This current note discusses and illustrates the conditions under which nonconvergence occurs in log-linked count models.

 

Talks and Tutorials

 
file

Midwest SAS Users Group (MWSUG)

Chicago, IL

September 29 – October 1, 2019

 

South Central SAS Users Group (SCSUG)

Baton Rouge, LA

October 17–18, 2019

 

file

SouthEast SAS Users Group (SESUG)

Williamsburg, VA

October 20–22, 2019

 

 

 

Quick Links

 
 

SAS® Statistics and Operations Research News

Don't miss important updates from SAS! Please add sas.com as a domain in your safe sender list.
Want to get more out of your relationship with SAS? Update your SAS profile
.

About this e-mail:
If you do not wish to receive future SAS® Statistics and Operations Research News editions, please unsubscribe.
SAS places great value in fair information practices and in connection with the management of our contact database, we would like to remind you that SAS Institute Inc. may use your personal contact details for marketing purposes, as stated in its privacy policy. To contact SAS via postal mail: SAS, SAS Campus Drive, Cary NC 27513 USA. ATTN: Legal Division/Privacy Manager.

SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ® indicates USA registration. Other brand and product names are trademarks of their respective companies. Copyright © SAS Institute Inc. All rights reserved.