SAS Statistics and Operations Research News

Posted: 11/11/2018

View as a web page


October 31, 2018

SAS The Power to Know logo



Just back from SESUG in St. Petersburg, Florida, and while the temperatures at RDU were about 40 degrees cooler than what I left at the Tampa airport, the warm fuzzies are still there from the conference! I love these regional events, where the smaller numbers make it more intimate and it’s possible to get to know many attendees. I had a great group for my tutorial on longitudinal data analysis and my other talks, and we’re already working on answers to questions posed about SAS/STAT® and other products. The organizers are already planning for next year in Williamsburg, Virginia, so if you live in the mid-Atlantic region, I would strongly recommend attending!

At headquarters, we are busy finalizing the next release of SAS® software, which will bring you additional features on the SAS/STAT and SAS/ETS® fronts. More about that next time! My colleagues are just back from Milan and the European Analytics Experience conference, where they touted numerous analytical products on the SAS® Viya® platform. If any readers attended, I’ve love to hear your thoughts.

Just a quick note: if you get the word that your paper has been accepted for SAS® Global Forum 2019, and you are relatively new at the paper-writing and presentation business and would like some guidance, there’s a very popular mentoring program that connects writers and experienced SAS Global Forum presenters. 

Here’s to whatever writing is on your list for this fall and winter!


Senior R&D Director, Statistical Applications


Technical Papers



Navigating the Analytics Life Cycle with SAS® Visual Data Mining and Machine Learning on SAS Viya

Extracting knowledge from data to enable better business decisions is not a single step. It is an iterative life cycle that incorporates data ingestion and preparation, interactive exploration, application of algorithms and techniques for gaining insight and building predictive models, and deployment of models for assessing new observations. The latest release of SAS Visual Data Mining and Machine Learning on SAS Viya accommodates each of these phases in a coordinated fashion with seamless transitions and common data usage. An intelligent process flow (pipeline) experience is provided to automatically chain together powerful machine learning methods for common tasks such as feature engineering, model training, ensembling, and model assessment and comparison. Ultimate flexibility is offered through incorporation of SAS code into the pipeline, and collaboration with teammates is accomplished using reusable nodes and pipelines. This paper provides an in-depth look at all that this solution has to offer.


Fitting Compartment Models Using PROC NLMIXED

The CMPTMODEL statement is a new enhancement to the NLMIXED procedure in SAS/STAT 14.3. This statement enables you to fit a large class of pharmacokinetics (PK) models, including one-, two-, and three-compartment models, with intravenous (bolus and infusion) and extravascular (oral) types of drug administration. The CMPTMODEL statement also supports multiple dosages and PK models that have various parameterizations. This paper introduces the new statement and illustrates its usage through examples. Related concepts are also discussed, such as the %PKCONVRT autocall macro (which converts PK data sets that are stored according to industry standard to data sets that can be directly used by PROC NLMIXED), extension to Emax models, prediction, visualization, and fitting Bayesian PK models (by using the MCMC procedure).



Economic Capital Modeling with SAS® Econometrics

A statistical approach to developing an economic capital model requires estimation of the probability distribution model of the aggregate loss that an organization expects to see in a particular time period. A well-developed economic capital model not only helps your enterprise comply with industry regulations but also helps it assess and manage risks. A loss distribution approach decomposes the aggregate loss of each line of business into frequency and severity of individual loss events, builds separate distribution models for frequency and severity, and then combines the two distribution models to estimate the distribution of the aggregate loss for each business line. The final step estimates a copula-based model of dependencies among the business lines and combines simulations from the dependency model with the aggregate loss models of individual business lines. This process yields an empirical estimate of the enterprise-wide loss distribution, which helps you develop an economic capital model. The process is characterized by both big data and big computation problems. This paper describes how SAS Econometrics software can help you take advantage of the distributed computing environment of SAS Viya to implement each step of the process efficiently.


SAS® Visual Forecasting: A Cloud-Based Time Series Analysis and Forecasting System

SAS Visual Forecasting, based on SAS Viya, is the next-generation SAS product for forecasting. It provides a new resilient, distributed, scripting environment for cloud computing that provides time series analysis, automatic forecast model generation, automatic variable and event selection, and automatic model selection. SAS Visual Forecasting features a new graphical interface that is centered on the use of pipelines; a new microservices-based architecture; and a new fast, scalable, and elastic in-memory server environment based on SAS® Cloud Analytic Services (CAS). It provides end-to-end capabilities to explore and prepare data, apply various modeling strategies, compare forecasts, override statistical forecasts, and visualize results. The workflow framework for model generation and forecasting is shared with SAS Visual Data Mining and Machine Learning and SAS® Visual Text Analytics. Forecast analysts and data scientists can also access the power of SAS Visual Forecasting through a flexible and powerful programming environment.



Technical Highlights


blog icon

Graphically Speaking

Distinguished Research Statistician Developer Warren Kuhfeld discusses marginal model plots, which display the marginal relationship between the response and each predictor in a regression model. Prompted by a question from a SAS user at the MWSUG conference this fall, Kuhfeld researched and here lays out how to build an ODS graph in layers.




Video camera icon

Lights, Camera, Action!

This newsletter brings a bonanza of new videos to you, from feature engineering to regime-switching regression to fitting hierarchical nonlinear mixed models. Get some popcorn and enjoy!

Executing Open Source Code in SAS Visual Data Mining and Machine Learning Pipelines

Feature Engineering in SAS Visual Data Mining and Machine Learning

Taming Text Data with SAS® Studio Tasks

Perform Power and Sample Size Analysis with SAS Studio Tasks

PROC OPTMODEL: Solving Optimization Problems with Hybrid Approaches

Regime-Switching Regression Using the HMM Procedure

Fitting Hierarchical Nonlinear Mixed Models Using PROC NLMIXED

Simulate and Save Critical Data with SAS® Simulation Studio


blog icon

The DO Loop

Distinguished Research Statistician Developer Rick Wicklin teaches you about optimization with nonlinear constraints in SAS, kernel regression in SAS, and radial basis functions and Gaussian kernels in SAS. 



Tech Support Points Out


Tech Support Points Out icon

Sample 63038: Predictive Margins and Average Marginal Effects

Predictive margins are estimates of the response mean and are typically used for fixing some, but not all, predictors in the model at specified values. The marginal effect of a continuous predictor at an observation is the instantaneous rate of change of the response mean at that point. The average of the marginal effects over the observations (AME) is often used as a measure of the effect of the continuous predictor on the response mean. A similar measure is the marginal effect estimated at the mean of the other predictors (MEM). For small samples, the AME is considered the better measure. A measure of the effect of a categorical predictor on the response mean can similarly be obtained as the difference in predictive margins at two of its levels. This is often considered the “marginal effect” of a binary categorical predictor.

The Margins macro fits the specified generalized linear or GEE model and estimates and tests predictive margins and marginal effects (AMEs and MEMs) for variables in the model. Differences and contrasts of predictive margins and marginal effects with confidence limits are also available. Margins and effects can be estimated at specified values of other model variables or at computed values such as means or medians.



Quick Links



Statistics & Operations Research Home
Bayesian Resources
SAS Analytics: 14.3, 14.2, 14.1, and 13.2
FASTats: Frequently Asked-For Statistics



SAS/STAT Procedures A-Z
Analytical SAS Software Video Portal
SAS/STAT Example Programs
SAS/ETS Example Programs



SAS® Statistics and Operations Research News








You Tube


SAS Blog


COMMUNITIES: SAS Statistical Procedures | SAS/IML | SAS/OR | subscribeunsubscribe