Spotfire: A Visual Experience
For quick analytics of large data sets, with specialized packages for chemistry, genomics and data mining, version 3.1 has much to recommend it
John A. Wass, Ph.D.
click to enlarge
Table 1. |
Tibco calls Spotfire 3.1 “predictive analytics accessible within Spotfire’s visual user experience.” I believe they are trying to convey a sense that large, complex data sets may be easily analyzed by both technical and non-technical staff, and the results shared with colleagues. This facilitates many “what-if” scenarios to gain further insights into what the data is telling them. Spotfire is very intuitive in that the developers have reduced the main menu and tool bars to their minimal essential functions, and this is enough, as further customization is possible in each of the main areas by right-clicking.
The software's main power is to quickly reduce large amounts of data to extract insight as to trends, patterns and relationships and to shed light on possible underlying processes. The program relies on graphics for the most part, as this is not statistical software, so the analytics are a bit more truncated. Table 1 gives a good overview of what the standard and express packages offer.
Importing data is mercifully simple. Excel spreadsheets are rapidly brought in with minimal clicks, quickly examined and converted to a program table automatically. From there, analytics may be performed from menus and/or toolbars.
Graphics
Graphics are the core function of Spotfire, and there is no scarcity of power in this area. Standard graphics include: bar charts, line charts, combination charts, pie charts, 2-D and 3-D scatterplots, map charts, heat maps, treemaps, parallel coordinate plots, box plots, and network graphs. An extremely useful tool, the filter panel will expand or truncate the data graph to display only certain segments that allow focus on important events or trends.
The software includes coloring, lasso and axis marking that facilitate analysis of patterns, clusters and correlations among sets of variables. The multi-scale bar charts and combination bar and line plots also allow for enhanced analysis of unstructured data to identify outliers and trends. By right-clicking on any graphic and selecting the ‘properties’ option, a wide variety of customization is possible, including marker size, shape and coloring.
Analytics
Here is where the analyst can generate numbers, p-values, r-squareds and correlations from simple tests. From the ‘Tools’ menu, one may select data relationships (linear regression, correlation, ANOVA, Kruskall-Wallis, and chi-square tests), k-means clustering, line similarity (comparing lines to a designated master line) and hierarchal clustering. Spotfire needs to ‘see’ certain formats for the data before it will perform the functions desired. For example, to get the k-means clustering and line similarity first requires the generation of line charts. This is not a major problem, however, as generation of the proper precursor is usually a simple operation.
Summary
Having previously used the software for several years, running through this version was like meeting an old friend. For quick analytics of large data sets, it has much to recommend it. Recognizing there are many technical areas where specialized analytics are required, the developers now offer packages for chemistry, genomics and data mining.
Spotfire 3.1 allows users to integrate S+ or R statistical scripts and models easily into an intuitive, fully interactive interface, making these powerful and widely used languages available through the program to further increase the flexibility of analyses. Non-statisticians can make use of the power of pre-written scripts to increase the sophistication of their predictive models. It is also possible to import and analyze data from various enterprise data sources, including SAP NetWeaver BI, SAP ERP, Salesforce.com, Siebel eBusiness Applications and Oracle E-Business Suite. Thus, the software is profitably used by both business and technical staff. View an excellent automated demo at stn.spotfire.com/gettingstarted/spotfire/300/en/Default.aspx?chapter=1
Availability
$79/month Spotfire Express
$399/month Spotfire for Analytics
Annual subscriptions available
TIBCO Software
212 Elm Street, Somerville, MA 02144
1-617-702-1600; Fax: 1-617-702-1700
mds@tibco.com; spotfire.tibco.com
John Wass is a statistician based in Chicago, IL. He may be reached at editor@ScientificComputing.com.
