Excel-based application of data visualization techniques for process monitoring in the forest products industry.
Techniques for data visualization can often provide important information and insight to forest products manufacturing process managers and operators as these techniques can be used to identify important relationships between various process parameters or significant properties of monitored characteristics. Many of the currently available visualization tools, however, tend to be fairly expensive and require significant programming expertise and are thus not easily accessible to some individuals or organizations. This paper describes data visualization and the potential for development of data visualization techniques that address the accessibility issue by using Microsoft Excel and Visual Basic for Applications to generate plots for monitoring or analyzing a data set or stream. Example data from a particleboard manufacturing process are used to illustrate the use of such an Excel-based visualization tool for process monitoring and analysis, and to discuss its impact on the decision-making process. The visualization tool is also applied to data from a study of the mechanical and physical properties of oriented strandboard panels, in order to illustrate its functionality for the quick display and initial analysis of experiment results.
Visualization can be informally defined as the process of transforming data to pictures (Schroeder et al. 1996) or the process of representing data as a visual image. The concept of visualization is not new. Cave drawings found in France are over 20,000 years old and the Chinese created the first known maps in the 12th century (Tegarden 1999). What is new related to the field of visualization is the increase in computing power, allowing collection of massive amounts of data and providing the power to efficiently execute computer algorithms to facilitate the extraction of information from the data sets, i.e., data mining.
Visualization techniques have been used to provide illustrations of physical characteristics of forest parameters such as tree height or forest density. Additionally, visualization through the use of maps and Geographic Information Systems (GIS) has seen wide application in forestry. Rautalin et al. (2001) illustrated the use of computer visualization to demonstrate the consequences of forest management alternatives to laymen and forest owners. Chertov et al. (2002) utilized visualization in the form of a sequence of interactive maps for forest growth simulation. Bishop and Karadaglis (1996) developed an application of advanced visualization techniques in combination with a geographic information system and linear programming. Their techniques were developed to facilitate prediction of environmental changes that can affect timber production, water catchment properties, recreational values, aesthetic values, energy usage, or employment opportunities.
Visualization has seen extensive application in scientific research and in highly complex manufacturing processes; however, the use of visualization for monitoring and studying the relationships and interactions of process parameters has seen limited use in forest products manufacturing. The current availability of computing power offers great potential for forest products manufacturing facilities in their efforts to improve their ability to monitor and analyze data from manufacturing processes. Manufacturers are constantly searching for an improved understanding of their processes. To meet these needs, several large software packages are currently available for visualization. In general, these packages are fairly complex and expensive and have been used primarily by larger companies, companies involved in highly complex manufacturing processes, and the scientific community.
Our research effort is aimed at illustrating the use of the existing platform of Microsoft Excel and Visual Basic for Applications (VBA), which is available on most desktop computers, to develop visualization tools that could be used by small and medium-sized manufacturers at a much lower cost. The combination of Excel and VBA provides a powerful basis for developing generic or custom visualization tools that can be used by many manufacturers who are already collecting process data in Excel. In this article, we will provide an overview of existing visualization software, a description of some graphs and diagrams that can be utilized to analyze and monitor the behavior of processes, and an illustration of the development of a VBA process monitoring tool for forest products manufacturers.
Visualization software tools
Advanced Visual Systems (AVS), IBM, The MathWorks, Inc., Research Systems, Inc., and SAS are examples of companies that currently provide major software tools for data analysis and visualization. We begin our discussion with a brief description of some of the commercial software packages offered by these companies. Summary descriptions of these and many other visualization packages are also provided by the University of Minnesota Supercomputing Institute for Digital Simulation and Advanced Computation (UMSI) (www.msi.umn.edu) and on the website for each individual company.
OpenViz is a visualization software package provided by AVS (www.avs.com) that allows the integration of data from any data source into a large variety of types of graphs and charts. The software allows companies to build two-dimensional and three-dimensional data visualization functions into data analysis and business intelligence applications (Davis 1999). IBM Visualization Data Explorer (www.ibm.com) is another visualization framework that provides a full set of tools for manipulating, transforming, processing, realizing, rendering, and animating data. Data Explorer allows for visualization and analysis methods based on points, lines, areas, volumes, images, or geometric primitives in any combination. The Interactive Data Language (IDL) from Research Systems, Inc. (http://rsinc.com/idl) is a widely used general-purpose visualization package that allows for the creation of 2-D and 3-D graphs with contours, isosurfaces, and slices. It includes tools for a "quick--look" interactive analysis and for large-scale commercial programming projects and it is heavily used for scientific data analysis (UMSI 1999).
MATLAB software (www. math-works.com), developed by The Math-Works, Inc., provides functions for visualizing both two- and three-dimensional scalar and vector data. Such functions exist for basic plots and graphs; three-dimensional plots; surface, mesh, and contour plots; and volume visualization. MATLAB also is frequently used in both the academic and scientific communities. SAS/SPECTRAVIEW (www.sas.com) is the SAS System's tool for data visualization and analysis. Geometric images representing multidimensional data can be created, analyzed, and modified. The software has been used in medical imaging, oil exploration, chemical analysis, and financial analysis. SAS/SPECTRAVIEW is interactive and works well with large data sets.
These visualization software packages typically require a fairly high level of programming or user expertise, or a significant financial investment, or both. The price and required expertise may make the use of these visualization software programs unrealistic for small or medium-sized manufacturers who do not require much of the functionality provided in these software packages. Consequently, we propose the development of a set of smaller scale visualization tools using Excel and VBA. An overview of some visualization tools that would facilitate improved process monitoring in a typical forest products manufacturing process follows.
Information graphics tools
Visualization is often used to identify any trends and patterns in data and to determine if those trends or patterns indicate an underlying relationship between two or more variables (Colet and Aaronson 1995). The ability to quickly evaluate and analyze the behavior of multiple process parameters, along with identifying or tracking relationships between process parameters, is essential to manufacturing process operators and managers. Many graphs, tables, maps, and diagrams can be used to analyze a data set. Two excellent references describing the plethora of graphic tools that exist are Information Graphics: A Comprehensive Illustrated Reference (Harris 1999) and Graphical Methods for Data Analysis (Chambers et al. 1983). Many of the tools discussed in these references, including scatter graphs, box plots, spider graphs, and radial plots, can be utilized to visualize process behavior and process parameter relationships in a forest products manufacturing facility.
The first of these tools, scatter graphs or scatter plots, enable the simultaneous display of either two or three different parameters. In particular, scatter graphs are often used to identify relationships between the parameters being displayed, such as the presence or absence of correlation. Because each instance of a set of parameter values is represented as a single point, however, scatter graphs are somewhat limited with respect to the amount of information that they are able to display.
Box plots, by comparison, are able to illustrate information about the distribution of individual parameters, including mean values, extreme values, and data distribution. Multiple box plots can be displayed simultaneously so that comparisons between different parameters can be made. Such comparisons of parameter or product properties could be useful when analyzing the differences between products produced on different machines, different processing lines, or different shifts. Differences in extreme values or data distribution might provide information relating to causes of differences or similarities in process or product parameters.
Spider graphs, otherwise known as radar graphs, are circular graphs used primarily to compare different series of parameter values. The value of each parameter is plotted on its own axis, and according to its own scale, and the points are then connected to form a polygon. Different data series may be compared by assessing the relative size and shape of these polygons (Harris 1999). Spider graphs are very effective when there is a clearly defined, and consistent, relationship between the size of the parameter values and the value or performance of the process that they represent. For example, they can be very useful in comparing two different paper products on the basis of their printing characteristics, where higher parameter values are more desirable in each instance.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
Spider graphs are less effective when the relationships between individual parameters are very complex and the significance of a polygon's shape becomes less clear. In such situations, a related type of graph, called a radial plot, that includes the individual parameter values but not a representation of the polygon may be simpler and more generally useful. A radial plot can provide powerful information about the behavior of multiple process parameters and the interactions between them. These capabilities are very useful in most forest products manufacturing processes, where the combination of a high number of correlated process parameters typically results in a final product with varying levels of strength and quality. We have chosen to utilize radial plots as the tool to illustrate the development of an Excel-based visualization package because of this simplicity and flexibility in interpretation, and because they are applicable to a wide variety of different problem situations. Details of our implementation of radial plots for process parameter visualization, including their application to the analysis of a time series and to statistical process control, are provided below.
Radial plots for process control
A radial plot is the equivalent of a rectangular column graph wrapped into a circle (Harris 1999). The horizontal axis of a rectangular graph is equivalent to the circular axis of a radial plot and the vertical axis of a rectangular graph is the radial axis of a radial plot. Figure 1 illustrates a basic radial plot of five process parameters where each parameter is represented by an individual ray within the plot. By placing the radial plot within the context of the underlying rectangular axes, we can also utilize the center point of the plot to represent the value of a corresponding response variable.
For example, suppose that we would like to represent data from a wood chip refining operation. The rays of an associated radial plot might represent: chip moisture content (MC) (var 1), specific energy (var 2), percent of wood type 1 (var 3), percent of wood type 2 (var 4). and percent of wood type 3 (var 5). The length of each ray represents the current value of its associated process parameter, the color of each ray can be used to signal whether the current value is within a target range, and the height of the plot's center point on the vertical axis is then used to represent the freeness of the resulting pulp (the rate at which water will drain through the pulp).
Although the precise value of each parameter cannot easily be determined from a single basic radial plot, the relative values of the parameters associated with different observations can be obtained visually by displaying a time series of radial plots (Fig. 2). Viewing a series of radial plots that represent the process behavior over the last several hours allows a process operator to identify trends and patterns in the various parameters, as well as to evaluate the impact of those patterns and trends on the response variables. For example, in examining Figure 2, the process operator can make several observations:
1. As chip moisture increases, freeness of the resulting pulp tends to decrease;
2. As specific energy increases, freeness of the resulting pulp decreases;
3. As the percentage of wood type 2 increases, freeness of the resulting pulp has a tendency to decrease.
These observations could not easily be made by simultaneously viewing multiple time series plots of the parameter values alone.
A multivariate control chart (MVCC) is another example of an approach that does provide this ability to simultaneously monitor more than one process variable. A variety of MVCCs have been introduced since the mid-1950s. Some of the more popular charts are Hotelling's [T.sup.2] control chart, which is used to detect shifts in the mean or covariance between several related parameters; the Squared Prediction Error (SPE) chart, which is based on the error between the raw data and a Principal Component Analysis (PCA) model fitted to that data; and Contribution Charts, which are used to determine the contributions of the process variables to either the principal component or the SPE for a given sample. For more information on MVCCs, the reader is referred to the following: Alt (1985), Murphy (1987), Montgomery (1991), and MacGregor and Kourti (1995).
The viewing of multiple radial plots offers an alternative approach when the development of MVCCs is not feasible, such as when there are a very large number of related process parameters. The two approaches can also be used together, so that additional parameter behavior can be monitored simultaneously with the charted parameters.
In order to capitalize on their potential for displaying system characteristics, we may augment a time series of radial plots by calculating the individual empirical distributions of each of the independent variables, over time, and then color-coding each individual ray to reflect the extent to which its value differs from the mean of its corresponding distribution. By using different colors to represent the observations of those auxiliary variables that differ from their mean value by pre-specified amounts, it becomes possible to visually identify potential relationships between variables that might not otherwise be immediately apparent. The idea of displaying several variables in a "control chart type setting" was motivated by Chernoff (1973), who introduced the concept of faces for displaying multivariate data.
The color scheme employed in the radial plot program used in this research assigns the color black to observations that are within two standard deviations of their mean, the color yellow to observations that are between two and three standard deviations of their mean, and the color red to all observations that differ from their mean by greater than three standard deviations. These deviations are calculated a posteriori, as full knowledge of the observations over the entire data set is available. They could also be calculated on a real-time basis, however, using all observations currently available. For expository purposes within this paper, we substituted a dotted line to indicate a parameter (a given ray in a radial plot) that is colored red in the actual application and a dot-dashed line to represent a parameter that is colored yellow. Solid lines are used to represent parameters that are colored black.
Radial plot generation using Microsoft Excel
Nottingham et al. (2001) provides several examples of the application of radial plots to problems of statistical process control. Color-coding of individual rays is used in those examples to identify when the auxiliary variable values are out of control, and to help determine the effect that such variation has on the value of the response. The radial plots in Nottingham et al. (2001) were generated using SAS System for Windows (SAS 1998), and the results were displayed within the context of that software environment.
In contrast, this Excel-based application of the radial plots not only simplifies the generation of the plots, but also dramatically improves the user's ability to manipulate the data being studied. Incorporating the use of Microsoft's VBA programming language into the Excel application allows for the generation of an extremely user-friendly interface to the system, and helps make it accessible to a wide audience. Since many practitioners in the workplace already use Microsoft Excel, and since VBA is automatically bundled with Excel in a standard installation, the Excel-based radial plot system described here requires neither new software nor additional training to be useful and effective. A good general overview of the capabilities inherent in the combination of Microsoft Excel and VBA may be found in Wells and Harshbarger (1997).
In particular, it is the built-in graphing capabilities of Excel, together with its capacity for data storage and retrieval, which make it an effective tool for dynamically generating radial plots based on a series of data observations. The addition of VBA allows not only for the development of userforms for a more user-friendly interface, but also allows for more detailed design of the radial plots themselves and the automation of many aspects of the process by which they are generated. Furthermore, Microsoft's Active X and OLE technologies, as built into VBA, provide the ability for an Excel spreadsheet to dynamically exchange real-time data with a variety of different ActiveX-compliant applications, including external databases, modeling packages, and even systems generating real-time PLC process data (Automated Solutions, Inc. 2002).
Interactivity between the user and the system is an important feature of any system that is used for data analysis, and the combination of Excel and VBA fully supports the ability of the user to both examine and manipulate the data in a variety of different ways. For those who are interested, a more detailed discussion of the VBA programming within the system may be found in Zobel et al. (2002). In addition, the Microsoft Excel spreadsheet containing the code for the radial plots displayed in this paper is available via e-mail from the first author (firstname.lastname@example.org).
Manufacturing process control example
Data input and interface
One of the distinct advantages of using Microsoft Excel as the platform for a data visualization and analysis tool is the ease with which the data to be analyzed can be entered or imported into an Excel spreadsheet (Fig. 3). In Nottingham et al. (2001), radial plots are produced that represent a process operating in a particleboard manufacturing facility. Those same data are utilized as an example application for the Excel-VBA radial plot system. The data consist of particleboard process parameter values of core and face resin content, core bulk density, and core and face infeed MC, along with the resulting response values of internal bond (IB).
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
The interface to the system provides cells within which the user may enter the names and values for each of the auxiliary variables (columns C through H, Fig. 3), as well as for the response variable (column I, Fig. 3), and then labels for the individual observations (column J, Fig. 3). Either alphanumeric labels (such as names) or numeric values (such as observation time) may be used for this purpose. A custom command bar (see top of spreadsheet, Fig. 3) will then allow the user to navigate through the different parts of the system. The command bar buttons allow the user to view a series of radial plots (View Radial Plot Series: Fig. 4), a single radial plot associated with a specified observation (View Single Radial Plot: Fig. 5), or a line graph that simultaneously displays, for each observation, the value of a chosen auxiliary variable (View Single Factor Comparison Plot: Fig. 6).
Allowing manual data entry into a spreadsheet application, as described above, can potentially lead to data validation and quality issues. Fortunately, Excel has built-in data validation functionality that can be significantly augmented through the use of VBA. The effectiveness of including such functionality, and the ease of implementation, depends on the intended use for the visualization system. If the system is always used to enable visualization of a specific type of problem, it can be quite straightforward to incorporate consistent data validation functionality. If, however, the system is used to analyze several different types of problems with differing data characteristics, the incorporation of automatic data validation functionality can become very complex.
Radial plot series module
The radial plot series module (Fig. 4) allows the user to dynamically select and display an ordered collection of radial plots associated with a subset of observations from the data set. Both the View Radial Plot Series button (on the command bar) and the Graph Options button (lower left corner of Fig. 4) present the user with the Graphing Options dialog box. As shown in Figure 7, this not only gives the observations (if any) that are currently displayed in the radial plot series, but also allows the user to choose new observations to display. These new observations may be specified either as a continuous range of data (Figs. 7 and 8), or as a specific, nonconsecutive, subset of individual data points (Fig. 9). With respect to the particleboard data set, this allows the user to visually interpret the differences in parameter values and behavior during different time ranges, facilitating, for example, the comparison of the process when two different types of resin are in use.
[FIGURE 5 OMITTED]
[FIGURE 6 OMITTED]
[FIGURE 7 OMITTED]
The Graphing Options box also allows the user to customize the radial plot series graph by changing the number of tick marks on the vertical axis (response), specifying the minimum and maximum response values to be displayed, and showing or hiding the gridlines. As shown in Figure 4, the user also has the option to print the radial plot series graph, and to display a legend identifying which radius corresponds to which auxiliary variable. If the user has chosen to view a continuous range of data within the radial plot series module, the Next and Previous buttons at the bottom of the graph will also allow the user to scroll forward and backward through the observations in the data set.
The series of radial plots displayed in Figure 4 can be used to identify which process parameters are contributing to the changes in IB. For example, it is likely that the low value of IB for the sample at observation 108 is at least partially due to the low value of core resin content, as illustrated by the short length of the ray representing core resin content. Similarly, the higher values of IB occurring at observations 101, 104, and 106 are likely being impacted by the higher values of core resin content, as illustrated by the longer rays representing core resin.
The radial plot series also allows the impact of multiple parameter values to be analyzed and compared simultaneously. For example, it may be difficult to determine the cause of the low value of IB when looking only at the values at observation 103. Core bulk density and core resin treatment are the parameters expected to most impact IB; however, the values of those parameters at observation 103 are similar to those that resulted in higher IB values at other times, such as observation 102. Additional information about interactions between all parameters must then be gleaned by analyzing and comparing the parameter values at different times, in this case observation 99 and observation 103. The value of core bulk density at observation 99 is higher than at 103 (as denoted by the length of the rays), as is the value of IB. When comparing the two points, one is able to determine that the higher IB at observation 99 occurred with lower MC values. The combination of high core bulk density and lower MCs resulted in a higher IB value than that observed at observation 103. The ability of a process operator to verify suspected relationships between process parameters and product quality allows the development of increased understanding on the part of process operators and managers, and can thus support subsequent analytical efforts to determine optimal combinations of the parameters.
[FIGURE 8 OMITTED]
[FIGURE 9 OMITTED]
Single radial plot module
As its name suggests, the single radial plot module (Fig. 5) allows the user to view the radial plot associated with a single specified observation. This functionality may be accessed in one of two ways: 1) from within the radial plot series module (Fig. 4) by clicking on one of the individual radial plots; or 2) by clicking on the command bar's View Single Radial Plot button from anywhere within the system. If the latter approach is chosen, the user is provided with a drop down box from which to select the specific radial plot to display. To change the selected observation, the user may simply click on the View Single Radial Plot button again and select a new observation.
Upon choosing a specific observation, the user is presented not only with the associated radial plot but also with the radial plot's identifier (label of the observation number) and the value of its response variable. The current value of each auxiliary variable, its associated mean and standard deviation, and its associated minimum and maximum value are also displayed, and each such set of information is color-coded to match the ray with which it is associated within the plot (Fig. 5). When the user clicks on a particular ray, the data associated with that ray are highlighted and the user is given the option either to view the data set (Fig. 3) or to view the single factor comparison plot associated with that variable (Fig. 6).
Single factor comparison plot
The user can also choose to view a single factor comparison plot, either by clicking on the View Single Factor Comparison Plot button in the command bar, or by selecting a particular ray in an individual radial plot and choosing to view its associated comparison plot. The former approach will prompt the user to choose the auxiliary variable to be displayed and will then display each of the values within the data set for that particular variable. The latter approach, however, will not only display the comparison plot for the variable represented by the chosen ray, it will also highlight the location of the current observation (from the individual radial plot) within the comparative plot (Fig. 6). For example, the data displayed in Figure 6 represent face MC entering the blender, with a diamond indicating the data point that represents the observation from the individual radial plot (observation 51). If the user had chosen to view the comparison plot by selecting the command bar button, the same data points would be displayed but no individual observation would be highlighted, since none would have been chosen.
Data analysis example
In addition to process monitoring, radial plots can also be used for analysis of experimental data. The plots can be used to quickly visualize the results of experimentation, as a precursor to statistical analysis techniques. To illustrate this use, we now look at an example where radial plots are developed using data from a study where the mechanical and physical properties of clear wood specimens and oriented strandboard (OSB) panels were characterized (Peters et al. 2002). One objective of the study was to characterize physical and mechanical properties of OSB manufactured from the various types of hybrid poplar clones. Different hybrid poplar wood was used to fabricate OSB panels, which were tested to determine flexural properties, internal bond, density, water absorption, and thickness swell. Visualization with radial plots provides opportunities to assess the similarities and differences of the physical and mechanical properties of the OSB manufactured from different hybrid poplar clones. Radial plot series (Figs. 10 and 11) were developed to display the bending properties and specific gravity of various hybrid clones. A comparison of the modulus of rupture (MOR) values for the sapwood and heartwood of the various clones is easily made using Figures 10 and 11. One can quickly identify which clones have the higher MOR values and evaluate the relationships between MOR, overdry specific gravity, air dry specific gravity, and modulus of elasticity using the radial plots.
The MOR of clone 184-411 is low in comparison to the other clones for both heartwood and sapwood, while the MOR for clone 50-197 sapwood is significantly higher than the heartwood. This observation might guide further analysis as these graphs provide a quick and efficient way to visualize results and look for relationships before statistical analysis begins. Additionally, these types of displays provide a process operator who is not versed in statistical analysis a tool to visually analyze the possible relationships between parameters.
[FIGURE 10 OMITTED]
[FIGURE 11 OMITTED]
This paper discusses a simple, yet powerful, system for visualizing complex, multivariate data sets. The system has been developed using a combination of Microsoft Excel and VBA. The visualization system incorporates a radial plot representation of the data, which utilizes color coding of the rays associated with auxiliary variables in order to identify their relative relationships both to each other and to a chosen response variable. It enables easy data entry and specification, and flexibility with respect to the type and scope of observations to be displayed. In particular, a radial plot for a given observation can be viewed, as well as a series of user-specified radial plots, and an individual comparison plot for any of the chosen auxiliary variables.
The system can be used for various types of analyses, including preliminary data analysis prior to statistical or analytical model development. It can also function as an on-line process monitoring system, as it effectively includes statistical process control (SPC) information (by using the color schemes for the rays of the radial plots as control limits) and allows for the monitoring of multiple parameters simultaneously. Also, radial plots can also be used simultaneously with multivariate control charts. For example, if Hotelling's [T.sup.2] statistic indicates that the process is out of control at time t, one can examine the radial plot at time t to determine which of the process variables has contributed to the process change. Regardless of the area of application, the choice of development environment allows for minimal expense and easy accessibility for most organizations and individuals, in stark contrast to many of the visualization tools currently available on the market.
[c]Forest Products Society 2004.
Forest Prod. J. 54(5):57-65.
Automated Solutions, Inc. 2002. www.automatedsolutions.com/products/commab.asp.
Alt, F.B. 1985. Multivariate quality control. In: Encyclopedia of Statistical Sciences. Vol 6. N.L. Johnson and S. Kotz, eds. John Wiley and Sons, Inc., New York.
Bishop, I.D. and C. Karadaglis. 1996. Use of interactive immersive visualization techniques for natural resources management. In: Proc. SPIE-The International Society for Optical Engineering. Soc. of Photo-Optical Instrumentation Engineers, Bellingham, WA. 2656: 128-139.
Chambers, J.M., W.S. Cleveland, B. Kleiner, and P.A. Tukey. 1983. Graphical Methods for Data Analysis. Statistics/Probability Series. Wadsworth & Brooks/Cole, Pacific Grove, CA.
Chernoff, H. 1973. The use of faces to represent points in k-dimensional space graphically. J. of Am. Statistical Assoc. 68:361-368.
Chertov, O., A. Komarov, G. Andrienko, N. Andrienko, and P. Gatalski. 2002. Integrating forest simulation models and spatial temporal interactive visualization for decision making of landscape level. Ecological Modelling 148(1):47-65.
Colet, E. and D. Aaronson. 1995. Visualization of multivariate data: Human-factors considerations. Behavior Research Methods, Instruments, and Computers 27(2):257-263.
Davis, B. 1999. Tools help visualize data. Information Week 748:36.
Harris, R.L. 1999. Information Graphics: A Comprehensive Illustrated Reference. Oxford University Press, New York.
MacGregor, J.F. and T. Kourti. 1995. Statistical process control of multivariate processes. Control Engineering Practice 3(3):403-414.
Montgomery, D.C. 1991. Introduction to Statistical Quality Control. 2nd ed. John Wiley and Sons, Inc., New York.
Murphy, J. 1987. Selecting out of control variables with the [T.sup.2] multivariate quality control procedure. The Statistician 36(5):571-581.
Nottingham, Q.J., D.F. Cook, and C.W. Zobel. 2001. Visualization of multivariate data with radial plots using SAS. Computers and Industrial Engineering 41(1):17-35.
Peters, J.J., D.A. Bender, M.P. Wolcott, and J.D. Johnson. 2002. Selected properties of hybrid poplar clear wood and composite panels. Forest Prod. J. 52(5):45-54.
Rautalin, M., J. Uusitalo, and T. Pukkala. 2001. Title estimation of tree stand characteristics through computer visualization. Source Landscape & Urban Planning 53(1-4):85-94.
SAS Institute Inc. 1988. SAS System for Windows, Version 7. 1998. SAS, Cary, NC.
Schroeder, W., K. Martin, and B. Lorensen. 1996. The Visualization Toolkit: An Object-Oriented Approach to 3D Graphics. Prentice Hall, Upper Saddle River, NJ.
Tegarden, D.P. 1999. Business information visualization. In: Communications of the Association for Information Systems. Vol. 1. Paper 4. http://cais.isworld.org.
University of Minnesota Supercomputing Institute for Digital Simulation and Advanced Computation (UMSI). 1999. User support--scientific visualization. www.msi.umn.edu/user_support/scivis/scivis-list.html.
Wells, E. and S. Harshbarger. 1997. Microsoft Excel 97 Developer's Handbook. Microsoft Press, Redmond, WA.
Zobel, C.W., Q.J. Nottingham, D.F. Cook, and D. Potts. 2002. Simultaneous analysis of manufacturing process parameters using Excel-based radial plots. Working Paper 02-091. Dept. of Business Information Tech., Virginia Tech, Blacksburg, VA.
The authors are, respectively, Associate Professor, Assistant Professor, and Associate Professor, Dept. of Business Information Technology, 1007 Pamplin Hall, Virginia Tech, Blacksburg, VA 24061-0235. This research was supported by a grant from Burlington Industries. This paper was received for publication in September 2002. Article No. 9546.
|Printer friendly Cite/link Email Feedback|
|Author:||Cook, Deborah F.; Zobel, Christopher W.; Nottingham, Quinton J.|
|Publication:||Forest Products Journal|
|Date:||May 1, 2004|
|Previous Article:||Kraft pulping response and paper properties of wood from densely stocked small-diameter stands.|
|Next Article:||Fuzzy knowledge-based modeling and statistical regression in abrasive wood machining.|