StATS: S-plus version 7 (April 19, 2005)
This page is moving to a new website.
I attended a web seminar by David Smith introducing the latest version of S-plus, version 7.This software, produced by Insightful Corporation, is one of my favorite products for producing advanced statistical analyses. Dr. Smith cited four advantages of Insightful software products:
He mentioned how Pillsbury Bakery had produced a web based statistical application to allow non-statisticians to review quality reports and charts. Ratings managers at Lipper (Reuters) use the rapid prototyping to react faster to new mutual funds introduced to the market. Eli Lilly uses custom reporting and visualization tools to assist with high throughput screening. At Kraft, product quality managers use these graphics to visualize sensory panel data to adjust recipes for consistency and aroma. At Verizon, managers process massive amounts of network performance data to optimize their networks.
The new features of S-plus Server include
A new library by Chao, Pinheiro, et al, allows for analyses of mixed models with non-Gasussian responses, such as repeated measures of counts and proportions with random effects. A library by Terry Thernau allows you to build mixed effects Cox models. This allows you to include random effects in a survival data model.
The S-plus workbench provides an integrated development environment of S programming teams. It includes a smart code editor that recognizes the particulars of the language and offers appropriate syntax highlighting and indenting. The code editor has an outline function which allows you to jump to various segments of your code and a history window that lists the previously submitted commands. You can integrate task within the code by including comments in the code itself. These tasks are highlighted in a special window and allow you to jump to the appropriate code. This workbench is based on the industry standard Eclipse framework.
S-plus now has a pipeline architecture for scalable data analysis. You can now manipulate and analyze data sets in the gigabyte size range. Previous verison of S-plus stored all of the data in memory, the new system stores data on the disk. S-plus has a new data type, the bdFrame (Big Data Frame). There are variants for time series and vector data.
The bdFrame uses the same tools as other S-plus objects such as the ability to use the square brackets to selects portions of the data frame. The bdFrame has news data manipulation functions for appending and merging and filtering data sets. S-plus has the ability to use SQL code for data manipulation. For large data sets, S-plus has the ability to produce hex-bin charts rather than scatterplots. With large data sets, a scatterplot often looks like a big blob of ink that is difficult to interpret.
S-plus has new out-of-memory algorithms for linear, logistics, and Poisson regression. You can apply arbitrary S-plus functions to data blocks using the bd.by.group() function.
Dr. Smith ended with a case study using US 2000 Census data. To read the data in as a bdFrame, the code looks like
You can remove rows with zero population totals using standard S subscripting techniques
but this requires two passes through the data. A new and faster alternative is
You can also use regular expressions to filter this data.
There are new "Big Vector" data types: bdNumeric, bdFactor, bdCharacter, bdLogical, bdTimeDate. These data types have the same efficiency advantages of the bdFrame.
There are two white papers:
A new course, S-plus 7 - Working With Big Data, is also available.
This page was written by Steve Simon while working at Children's Mercy Hospital. Although I do not hold the copyright for this material, I am reproducing it here as a service, as it is no longer available on the Children's Mercy Hospital website. Need more information? I have a page with general help resources. You can also browse for pages similar to this one at Category: Statistical computing.