Boxplotted Tables and other Graphical Techniques for Exploratory Data Analysis.
Abstract
This thesis presents several interactive computer programs for the analysis of multivariate data. A special case is that of panel data; multiple times series of short length. The first program, BOXPLOTAB, handles this type of multivariate data; it is an enhancement on an existing graphical technique for exploratory data analysis known as BOXPLOTS. The program works by appending boxplots as column dividers in a table of the raw data which originates the box plots. This combination of the raw data and the graphical representation of that data improves the understanding of the characteristics of the data in exploratory and descriptive applications; differencing and tracing of data through the table is also implemented. This thesis also presents and explores the use of other graphical techniques for exploratory data analysis of multivariate data such as STAR plots, PROFILE plots, CODED SCATTER plots and CODED DRAFTMAN plots. These techniques are examined and implemented in a series of computer programs which produces these graphical displays. A technical description of each computer program is presented and user implementation procedures are discussed. The programs are implemented in APL and run in conjunction with the experimental IBM APL Graphics program GRAFSTAT. To demonstrate the use of these techniques, an analysis is conducted on several sets of multivariate data.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1986
- Accession Number
- ADA175146
Entities
People
- Juan M. Isusi
Organizations
- Naval Postgraduate School