JMP For: Analytical Application Development • Business Visualization • Design of Experiments • Exploratory Data Analysis • Interactive Data Mining • Modeling • Quality Improvement • Reliability • Statistics • Visual Six Sigma
JMP® for Exploratory Data Analysis
As you collect data from more and varied domains, and in much larger quantities, you may be scrutinizing your data for the first time. Exploratory data analysis (EDA) can help you find structure in all of this data.
If needed, EDA can also help guide you in building a useful model. Even if you have seen lots of similar data and expect the modeling to be routine, checking model plausibility and verifying assumptions via EDA remains an essential prior step.
By its nature, EDA is heuristic, open-ended and dynamic. It also often involves significant data quality and aggregation steps as you try different visualizations to let your data best tell its story. The interactive graphics and data management in JMP are ideal for EDA. And, even if you have significant data volumes, the in-memory architecture of JMP makes it responsive and fun to use, no matter where your data leads you.
- Data Selection and Management
- Linked Interactive Graphs
- Linked Interactive Analysis
Handling incongruent cases appropriately is an important step in EDA. Individual rows in a data table may be selected, colored, marked, labeled and excluded or hidden directly from any visualisation in which they are displayed, and such changes immediately propagate to all open displays. You can use Missing Data Pattern to quickly segregate cases that are incomplete, while Summary allows you to aggregate detail-level data into a linked table to allow visualizations at a higher level of granularity. The Data Filter can make all displays conditional on your selection of variables and their levels and ranges, as you make them. This allows you to rapidly review, characterize and appropriately handle all cases that satisfy the currently imposed condition. Cases can also be colored by variable using standard or custom themes.
Missing data pattern in three measured parameters, with linked displays showing the association with covariates and values.
Perception is personal, and the open-ended nature of EDA means that you will develop your own style of analysis. JMP provides a wide repertoire of visualizations so that there are few limitations. Various tools allow you to pan and probe these displays, and zoom in as required. The Graph Builder is a powerful and innovative platform that allows you to interactively build trellis displays with multiple x and y grouping variables and containing graphical segments such as bar charts, histograms, line charts and contour plots. And if the dimensionality of the data is high, you can use the Parallel Plot with coloring and transparency to reveal structure when there are many cases. But often insight comes from using multiple visualizations simultaneously, and JMP’s linking and Data Filter make this approach even more useful.
Using the Data Filter to obtain conditional selections by ‘biscuit_category’ in two linked displays of sales data colored by retailer.
With JMP you can be genuinely data-driven. In many cases you can peruse an initial exploratory analysis directly from the visualization itself, making choices that are informed by what you actually see rather than by what you expect. Typically, tabular output is appended directly to the same report window, and the display will be augmented by a visual representation of the analysis results (such as a regression line with confidence intervals). And, with the correct options set, you can make the analysis results instantly respond to the selections you make in the Data Filter.
Auto insurance claims data by age and colored by gender – Initial exploratory report, then augmented with output from the chosen analytical approach.
More resources for Exploratory Data Analysis
Demos
On-Demand Webcasts
Contact JMP Sales
877.594.6567 (US)
