TEAM Zarate Tarrani Notes From the Field

Last week I briefly discussed reliability and quality in a few entries. Both of these topics are heavily grounded in probability and statistics, as are most of of the activities in which IT professionals engage. I use three basic tools, depending on the type of work I'm doing: Excel for business and simple computational problems and analysis, MathCAD for more complex work, such as queuing and linear programming/optimization, and risk analysis, and STATVIEW for heavy statistics.

Each tool has its place. If you're using Excel and want to learn how to tap into its power I recommend Management Decision Making: Spreadsheet Modeling, Analysis, and Applications . This college-level text is also useful to business and IT professionals because it provides a refresher for decision techniques that are the foundation of a number of disciplines. The book is divided into two parts:

1. Deterministic methods, mainly focused on linear programming and optimization
2. Probability and queuing.

What makes this book valuable to the practicing professional is that it uses one of the most common business tools, Microsoft Excel, and shows how to apply this tool to real world problems. The accompanying CD ROM comes with TreePlan for developing decision trees and CrystalBall for Monte Carlo simulation, as well as workbooks that are used to support the plethora of realistic examples used throughout the book.

Although the book is business-oriented and better suited for operations analysis and MBA students and practitioners, I've used it as a reference for project planning, computer system capacity planning and performance analysis, and IT security risk management - these practical uses of the material show the value of the book in the real world. The supporting web site that the author maintains has materials for lesson plans, errata and additional resources that make this book particularly valuable for the academic and business environments.

For more intense computations the best tool is MAthCAD 2001 Professional. This progam is valuable because there are limitations to spreadsheets for performing advanced statistics, differential equations and graphing. Yes, if you're clever with common spreadsheet applications, such as Excel, you can work wonders. However it's time consuming, clumsy after a certain point, and often requires third-party add-ins. MathCAD, on the other hand, allows you to perform complex operations with a simple drag and drop from its extensive library of built n operators and functions.

I use it for computer systems capacity planning and performance analysis, general statistical analysis and probability in project planning and control. In that respect I haven't begun to tap into the power of this program because, especially calculus and matrix operations. However, what I do use it for gives me an idea of the time savings that results from building equations by dragging the symbols onto the screen, adding the variables and seeing the results immediately. The graphing function is as easy (and powerful). What I can do in MathCAD in less than a minute would take hours to set up in Excel, for example.

What I particularly like about MathCAD is the document management capabilities that are built in, the fact that it seamlessly integrates with Microsoft Office applications (you can drag your equations and graphs into a Word document, for example, with the same ease as an Excel table or graphic), and ability to save your documents as HTML.

Another strong point about MathCAD is the large collection of files and electronic books that are freely available from the publisher's website. They serve as a clearinghouse and solution sharing point for MathCAD users and the solutions that are available cover every business, scientific and technical discipline. An example that is in my technical area of expertise is the Closed queuing network analysis solution that came in handy when I was analyzing batch processing optimization.

If you work with equations and have reached the limits of your spreadsheet application you may find MathCAD to be a great value. You'll certainly become more efficient and productive with it. You'll also find that the learning curve is relatively flat because the user interface is similar to Microsoft's Office family of products. You'll probably wonder how you got along without the ability to build equations with drag and drop shortly after you begin using it. Technical support is responsive, the documentation is clear and complete, and the publisher's web site provides a wealth of add-ons and other tools.

When it comes to probability and statistics one of the best programs is Statview 5.0, which is one of the more popular statistical programs used in business and scientific applications. Amazon sells a student version, and the only difference between it and the professional version is the licensing. The student version has all of the features and capabilities, but you are restricted by legal terms and conditions of the license from using it outside of the academic environment. This is also an ethical issue.

What it contains: a comprehensive suite of descriptive statistical, statistical process control capabilities, regression, analysis of variance, factor analysis and non-parametric test functions are built in. The power and ease of use comes from the innovative user interface, data management and reporting features. In particular, the user interface stands out as my favorite feature because of its simplicity and power. It contains two types of windows, dataset and view, which accounts for its simplicity, and interactive browsers that allow you to access and data and apply analytical functions that account for the power. The dataset window is similar to a spreadsheet, and the plethora of functions allow you to perform any common (and many less common) analyses. You can also analyze more than one dataset in a single analysis, and you can save the work as a template to save time for similar analyses. Results can either be displayed in tabular or graphical format at literally a click of your mouse. The preview feature allows you play 'what if' in the same manner as spreadsheets, by changing data or parameters, seeing the changes. Page layout is powerful and flexible, which is something one expects from a professional tool.

If you are a student and qualify for the license terms and conditions STATVIEW is a bargain and a time saver. In addition to getting a tool that will make short work of statistical analyses regardless of whether you're majoring in social sciences, business or technology, you'll be learning the same software that you'll probably use after graduation because SAS Institute, the publisher, is one of the most respected names in statistical software.

If you are not a student and want a more focused statistical program I recommend JMP Statistical Discovery 4.0. Unlike STATVIEW, which is for general statistical analysis for a number of business and technical disciplines, this application is focused on operations analysis, statistical process control and design of experiments.

What makes this an industrial-strength tool is the fact that it works with mainstream applications, such as Microsoft Excel (open tables can directly access Excel files), and with any database that can be accessed via ODBC (MS Access, Oracle, SQL Server, etc.). Further ease of use is provided by the column browser that combines a familiar spreadsheet view with powerful navigation capabilities. Additional flexibility is given by extensive reporting features (easy layout, save as HTML for web publishing, and editing data while in the report function). If you want to automate repetitive tasks JMP also includes a scripting language that is reasonably easy to learn and is integrated with the formula editor.

One of the more powerful functions is design of experiments, which sets this application apart from general statistics programs - if you use DOE or Taguchi methods then you have ample justification for investing in JMP because you're be significantly more productive.

In addition to DOE/Taguchi methods JMP does descriptive statistics (eliminating the need for a separate statistics program), linear models, correlations and multivariate computations, statistical process control charts, and time series analysis. These capabilities make it ideal for anyone involved with quality assurance, R&D, operations analysis and reliability modeling. This is a professional tool that will save more time than more general packages such as the company's other product, STATVIEW, or applications such as MathCAD.