Outliers , Inliers , and Just Plain Liars-- New Graphical EDA + ( EDA Plus ) Techniques for Understanding Data
暂无分享,去创建一个
Graphs, the natural language of mankind, offer even novice data analysts a quantum leap in their data editing/analysis capability. In our dynamic, rapidly changing world, we are faced with a virtual flood of data-often from very vital/complex systems. Unfortunately, this flood of data can likewise be a real challenge to understand/analyze/edit. Many of our traditional/conventional data analysis/editing methods generate fixed-formula printouts requiring an analyst to review and correct the fields of records that are thought to be erroneous. There are many limitations to these conventional methods-even when well designed. They sometimes overlook basic methodological problems, they typically channel the reviewers in a manner that either may not allow a number of the errors to be found, or they focus on traditional (outdated?) relationships, and they often do not account for changes in the data. A special course has been designed by the author to teach a number of powerful graphical based methods to deal with these problems. This very popular course uses new, easy to learn (point-and-click), interactive Exploratory Data Analysis (EDA) software packages (SAS's JMPâ and Insightâ graphical data analysis software)-and makes these techniques very straightforward to apply. These graphical methods can first be applied in an exploratory manner-to discover nuances that conventional methods are likely to miss. Then, special interactive EDA + graphic forms (creating "live" graphs) can then be used in a straightforward, highly productive manner to edit these data.. (NOTE: The "+" is used in EDA+ is used to signify the enhanced interactive EDA methodology developed by the author to enhance EDA methodology-this is also called EDA PLUS.) Lastly, these new EDA+ graphical methods allow "inlier" detection and confirmatory review of data-to not only find hidden relationships within these data but to also assure that corrections made during to the editing process have worked well. NOTE: Although these interactive EDA techniques are also taught as part of a graduate level statistics course by the author, they offer even novice subject matter specialists a quantum leap in their data analysis capability. As such, it can be taught to a variety of audiences.
[1] A. F. Smith,et al. Statistical analysis of finite mixture distributions , 1986 .
[2] V. Hardman. Author Information , 2021, Disability and Health Journal.
[3] Peter L. Brooks,et al. Visualizing data , 1997 .
[4] John Sall,et al. Leverage Plots for General Linear Hypotheses , 1990 .
[5] HOW EXPLORATORY DATA ANALYSIS IS IMPROVING THE WAY WE COLLECT BUSINESS STATISTICS , .