Essential Regression and Experimental Design in MS Excel - free, user-friendly software package for doing multiple linear regression, step-wise regression,polynomial regression, model adequacy checking and experimental design in MS Excel
Essential Regression and Experimental Design
Dave Steppan, Joachim Werner and Bob Yeater
Copyright 1998 All Rights Reserved
Preview Program (View Quick Start Guide)
Essential Regression and Experimental Design for Chemists and Engineers was developed as an easy-to-use book with an accompanying software package which allows non-statisticians to analyze experimental designs and quantitative data using polynomial and multiple linear regression in a straightforward and understandable manner. From our experience as chemists and engineers, these two variations of regression analysis are the ones used most often to analyze data. They are the "essential" tools in data analysis. Recognizing the widespread use of Microsoft OfficeÒ software, we developed Essential Regression software as a MS ExcelÒ Add-In (compiled Excel Macro). The user can work in the familiar and powerful data analysis environment of Excel and does not have to learn a new statistical software package. Other benefits from working directly in Excel are that it trivializes some of the most time consuming steps of regression analysis when compared to large conventional packages because the entire input and output of the regression lies within a standard spreadsheet workbook which eliminates the need to learn a new interface. They include:
The book and software are intuitive and guide the reader through the process of setting up a regression model and analyzing it. The software also contains an on-line help file which contains thumbnail descriptions of the significance of the output of the regression analysis and detailed instructions on how to use the software. This help file is no substitute for reading and understanding the book.
The book and software describe and implement all the tools needed for a complete linear regression analysis. Up to about 20 independent variables or regressors can be selected in a multiple regression, and second and third order models (including interactions) can easily be set up using the built-in dialogs. In the Polynomial Regression module, up to ninth order polynomials can be constructed. There are limitations with respect to the number of data points. The accompanying software is best suited for small and intermediate data sets of 50 to several 100 data points. This is a size which most often occurs in "everyday problems" encountered by students and scientists. It was not developed to handle large data sets of several thousand and more data points used by, for example, sociologists or pharmacological researchers.
The following approach is repeated throughout the book. A theoretical discussion of a statistical technique is presented followed by chapters which explain the features of the software pertaining to the theory discussed before. The sequence in which the theory is introduced follows an order which is most likely employed by the user: introduction to regression and types of models, ANOVA, hypothesis testing, outlier analysis, and graphical evaluation including surface plots. At the end of the book, a tutorial is included with data sets (also included in the Excel spreadsheets which come with the software) which are analyzed to illustrate the utility of the software. All the analyses presented can readily be reproduced by the reader. The book starts with the usual discussion of coefficient of variation and ANOVA analysis. It contains a variety of sections on different statistical parameters and residual analyses useful for model adequacy checking. For example there are sections on stepwise regression ("auto fitting") techniques, the effect of response and factor transforming, and the detection of outlier, influence and leverage points. Although the treatment of linear regression is very complete, the book is not intended as a fundamental theoretical textbook of linear regression aimed at statisticians. It is intended to teach regression to non-statisticians by applying linear regression to real data sets.
Experimental design is covered as it relates directly to regression analysis. This restricts the design package to factors and responses that are continuous, quantitative variables. Screening designs including full and fractional (Resolution 3-5) 2 level factorial and Plackett-Burman designs are covered. Response surface modeling (RSM) designs including face centered, circumscribed and inscribed central composite designs and Box-Behnken designs are included. The advantages and disadvantages of the various design types are covered. Advanced ideas such as aliasing, orthogonality, rotatability and sequential experimentation are explained.
The software accompanying Essential Regression and Experimental Design for Chemists and Engineers delivers all the tools necessary for a thorough, complete experimental design and linear regression analysis combined with easy handling and impressive output possibilities which rival the features of much more expensive and much less intuitive statistics packages.
Even as we go forward toward an electronic society, traditional publishing media (books) show no signs of being dethroned as the way to learn detailed technical concepts. However, books with illustrative examples and software that can be immediately applied do represent a vast improvement over a solely traditional approach. We believe that this "learning by doing" approach, along with a reasonably complete fundamental treatment represents an ideal way to learn new and useful technology. This is especially true for well-known and well defined concepts such as regression. We hope that you find Essential Regression and Experimental Design for Chemists and Engineers a good example of this new hybrid type of book.
DDS, JW, RPY
Essential Regression - Free Statistical Software for Microsoft Excel.
This page hosted by Get your own Free Home Page