The DO Loop

Distinguished Research Statistician Developer Rick Wicklin explains the STDCOEF option in PROC GLIMMIX and how it differs from the STDB option in PROC REG, and he also discusses an outlier-detection method in time series analysis called the Hampel identifier. Learn about the Iman-Conover transformation and how to use it. Finally, Rick offers an introduction to simulating correlated data by using copulas.

The SAS Data Science Blog

Yongqiao Xiao discusses how to streamline the scoring process of multiple analytic stores, while Hardi Desai talks about how she and her team were able to successfully develop and deploy a physical distance monitoring solution using computer vision for a SAS customer. Javier Schloemann how he and his team used SAS Viya to solve the problem of intelligently searching for new images in a large database of old images. Ricky Tharrington follows up his first post, an introduction to parallel processing, with this post that discusses parallel processing with CASL, the scripting language of the CAS server. And finally, Fijoy Vadakkumpadan gives you a behind-the-scenes look at the loadImages action in SAS Viya.

Precision-Recall Curve for Imbalanced, Rare Event Data

The receiver operating characteristic (ROC) curve and the area beneath it (AUC-ROC) are perhaps the most commonly used graphic and statistic for assessing the performance of binary-response models or classifiers. However, as described by Saito and Rehmsmeier (2015) and others, the ROC curve and AUC-ROC can be misleading when the proportions of events and nonevents become very imbalanced. An example would be when there are very few observed events. They note that the statistics behind the ROC curve, sensitivity (also known as recall) and specificity, are invariant to the degree of imbalance. When interest focuses on the ability of the model to predict a high proportion of true events among the predicted events, known as the precision or positive predictive value (PPV), a plot involving that statistic can be more informative. The fact that the precision, unlike specificity, is sensitive to the degree of imbalance allows a plot of precision to more accurately reflect the measure of interest regardless of the degree of imbalance. This is the advantage of the precision-recall (PR) curve and the area under it (AUC-PR). The PRcurve macro plots precision against recall and computes the AUC-PR by using an appropriate interpolation method. Although linear interpolation can be used between points in ROC space, that is not the case in PR space. The macro can also find optimal points on the PR curve that have maximum values of the F score and Matthews correlation coefficient (MCC).

Ask the Expert: How Do I Authenticate to SAS Viya?

Tuesday, August 3 | 11:00 AM ET

To access SAS data, services, and APIs from outside SAS, you must first log in. This requires registering clients and creating access tokens to allow authentication to and authorization of SAS. Join this webinar to learn the various techniques to connect to SAS Viya. You will learn:

Register Now

Ask the Expert: How Do I Use a Hash Object in Place of the LAG Function?

Thursday, August 19 | 11:00 AM ET

Join this webinar to learn how the LAG function works, to get an introduction to the hash object, and to understand why it can be helpful with autoregressive time series forecasting. The webinar will provide:

Register Now

Ask the Expert: How Do I Use SASPy to Interface with SAS from My Python Code?

Tuesday, August 24 | 11:00 AM ET

Join this webinar for an introduction to SASPy, our open source Python interface to SAS. We’ll explore use cases, resources, and capabilities. The webinar will provide:

Register Now

JSM 2021

August 8−12, 2021
Virtual

SAS is a Gold Sponsor and will have a virtual EXPO booth.

ASA Biopharmaceutical Section Regulatory-Industry Statistics Workshop

September 21−24, 2021
Virtual

International Alliance for Biological Standardization (IABS): 7th Statistics Workshop

November 8−11, 2021
Virtual