Choose Language Hide Translation Bar
Coming in JMP 12: Selection Filtering

Interactivity has always been a key feature of JMP that enables discovery through both data exploration and report generation. Transform Columns keep you "in the flow" during initial exploration. Interactive features in reports, such as brushing and selection linking between tables and reports, allow you to continue to ask questions throughout the analysis. The Local Data Filter and Column Switcher streamline this process, allowing you to interactively subset or change analysis variables without having to recreate a report. JMP 12 introduces the Selection Filter, which further enhances the interactive report generation process by filtering one set of graphs or reports based on the selection performed in another set of graphs.

Data Filters 


Figure 1: A Local Data Filter controls the subset for a single report


Data filters are commonly used to explore subsets of data based on specified criteria. As shown in Figure 1, the Local Data Filter in JMP provides interfaces for filtering data based on the values in one or more columns. Using a Local Data Filter allows you to explore subsets of the data within a given report without affecting the source data table or other reports that are dependent on the same table. Common workflow scenarios often involve multiple analyses linked together. For example, you may want to see a distribution of values in a column before deciding which subset to analyze. In previous versions of JMP, this process could be performed manually by creating subset tables and multiple reports. The manual approach can lead to a large number of windows being created for data tables and reports, requiring more mental bookkeeping to understand the analysis. With JSL scripting, it would be possible to streamline the process and provide a more cohesive workflow presentation. Using Selection Filters in JMP 12, this type of workflow can be configured interactively within JMP and saved for future use and sharing with colleagues. 

Selection Filters



Figure 2: Selection in the Treemap determines the data to be analyzed in the Oneway report


Selection Filters are similar to Local Data Filters, but they use one or more graphs in place of the traditional filter interface. Rather than using selection linking to share the selection state with the data table and other reports, a Selection Filter uses a private selection state that is independent of the data table and other reports. This private selection is then used to determine which rows to include in the filtered graph. Figure 2 demonstrates the use of a treemap as a filter for a Oneway analysis. The treemap shows all of the data, while the Oneway report is filtered to show only the rows that are selected in the treemap. As in the case of the traditional data filter, if nothing is selected in the filter, then all of the data will be shown in the filtered platform.


Figure 3: Two filter graphs provide multiple way to select and visualize the subset of interest.

Figure 3: Two filter graphs provide multiple ways to select and visualize the subset of interest.


Selection Filters are not limited to using a single graph within the filter or a single report that is being filtered. Figure 3 illustrates the use of two platforms within the same Selection Filter. The map provides a way to select individual states, while the region histogram provides a way to select states by group. In this example, the two reports are linked to the same selection, so that selection in either graph is reflected in the other, while the filtered Graph Builder shows only the selected data. It is also possible to link selection filters hierarchically, creating a filter that successively narrows the subset of interest based on selection in multiple graphs.


Linking multiple platforms together using a Selection Filter provides a streamlined analysis flow -- you choose both how to select the data of interest and how to view or analyze the subset. The combination enhances the power of each platform. Consider the possibility of using a hierarchical clustering platform as a Selection Filter. By linking another platform to the selection in the cluster hierarchy, you can interactively create a customized application that displays meaningful cluster diagnostics.


In a future blog post, I will demonstrate the process for creating a Selection Filter in JMP 12. Please comment on combinations of graphs or analyses that you frequently use together or would like to see explored. The combinations and applications are vast. I look forward to hearing your ideas!


Editor's note: This post is part of a series of previews of JMP 12 written by the people who develop the software.

Article Labels

    There are no labels assigned to this post.

1 Comment

Michael Clayton wrote:

My favorite graph if the Variability Plot (called Multi-Vari plot in Minitab and other systems). The use of stratification to get Variance Components can only be done in this plot for nested or crossed multi-level factor studies...visually. But it is classed as gage study tool, when we actually use it for quick look at all DOE's when completed, prior to complex modeling, as well as in SPC studies where the same plots often include multiple tools and gages and recipes for some product parameter result.

Anything that helps us Pareto the many possible causes of variation will help.,,on any graph platform.

Also for EDA on complex test data (500 or so electrical test parameters per die within 25 wafers per batch across hundreds of lots, over time) the newer ideas on finding critical parameters to assign to improvement efforts, other than older Hypothesis Tests, would be very helpful. Alpha and Beta risks are tricky issues with "big data" in precision manufacturing operations often leading to wasted effort in DOE's.