Computer Programs for Epidemiologists–PEPI Version 4.0

Computer Programs for Epidemiologists–PEPI Version 4.0 J. H. Abramson and Paul Gahlinger. Salt Lake City, Utah: Sagebrush Press, 2001. ISBN 0-9703130-2-0. 305 pp, paper and CD-ROM, $59.95. “There is a great satisfaction in building good tools for other people to use.” Freeman Dyson Like me, many readers of this journal received their epidemiologic training during a bygone era, when data analyses required punching coded instructions on IBM cards for processing in a mainframe computer. We physically carried our cards to the computer, somewhere on campus. We considered ourselves fortunate to have the benefits of computing. (Never mind the frustrations of finding that our run had failed due to a syntax error in our instructions.) We were happy that we could probe data in ways that we never thought possible. With the replacement of card punchers by computer terminals, we could actually dispatch our instruction sets with a press of a button. Even so, epidemiologic computing was far from being either convenient or efficient. This state of affairs changed dramatically with the advent of personal programmable calculators and computers in the late 1970s. This created the impetus for epidemiologists to write specialized software for tasks that had required centralized computing resources. The history of personal computing in epidemiology has produced many unsung heroes. Probably the first landmark contribution was the compilation by Kenneth Rothman and John Boice in the late 1970s of a series of programs they had written for the HP-67 calculator. These programs performed analysis of crude and stratified data from case-control and cohort studies, sample size calculations, computation of binomial confidence intervals, and life table analysis. They became the methodologic staple of Epidemic Intelligence Service officers-in-training at the Centers for Disease Control in Atlanta. The initial compilation was expanded in 1982 with the adaptation of all the original programs to the legendary HP-41C programmable calculator. 1 Rothman and Boice’s formulas and algorithms inspired many epidemiologists throughout the world to write their own code and macros for various computing platforms. The early 1980s also saw the advent of spreadsheet programs for the then nascent IBM personal computer. The interconnected matrix arrangement of spreadsheet software was quickly recognized by epidemiologists as a handy tool, and it inspired many a weekend programmer to make a contribution to the profession. Three attributes have served as the driving force behind so many useful epidemiologic programming contributions in the past 25 years: talent, altruism, and passion for the hobby of writing computing code. Many epidemiology programmers have derived satisfaction from placing their software in the public domain and from learning that others have used it, validated it, and perhaps even improved on their algorithms. Epi-Info, the most well known of these software programs is the best example of a product of this generous attitude, particularly that of one Andrew Dean, of the CDC. 2 Another case in point is the collection of computer programs that are the focus of this review. Over a period of more than a decade, J. H. Abramson and Paul Gahlinger have dutifully written, adapted and perfected more than 40 programs that address a variety of epidemiologic study design and analysis needs, from sample size and power calculations to logistic regression and meta-analyses. As the authors aptly state in their introduction, they aimed to complement the features found in commercial software packages by filling the gaps in the biomedical data-analysis toolbox. This popular collection, named PEPI for Programs for EPIdemiologists, is now in its fourth major revision, after multiple updates and additions. It is a veritable “Swiss army knife” of utilities for epidemiologists and biomedical researchers. One will find here more analytic options for a simple 2x2 or 2xK table than will probably be needed during an entire epidemiology career. Virtually all common statistical procedures that require input of frequency data (matched or unmatched) are covered by the PEPI programs. Several utilities offer useful analyses of continuous data as well. The collection has tools for a variety of specific needs in parameter and interval estimation (including computationally intensive exact estimation), significance testing, goodness of fit tests, trend tests, standardization, life table analysis, and Poisson and logistic regression. There are also tools for study planning or overview, such as sample size and power calculations, random numbers, correction for misclassification, combination of measures of association, P values from specific distributions (normal, t- and F-distributions), and other useful utilities. Most programs are written in Pascal for the DOS operating system, which makes them small enough to be used in less endowed computers, including some handheld devices. The largest of these programs is barely more than 100 kilobytes. Nearly all run in Windows within a DOS “session.” Those weaned in epidemiologic computing during the Windows era will miss the graphical user interface and may be a little distressed to find that their hands will have to be on the keyboard, not the mouse. However, once one becomes familiar with these utilities these details should not be problematic. The programs are very simple to use and prompt the user for data input. A few require reading a previously prepared dataset, but easy instructions for such preparation are provided. Over the years, this reviewer has used PEPI programs numerous times and has found them to be extremely useful. They perform as they should. Abramson and Gahlinger have enlisted many colleagues to field test the PEPI utilities. All obvious errors and “bugs” have been eliminated during the multiple revisions to the Pascal code. A book accompanies the CD-ROM with the PEPI utilities. Each chapter is devoted to a single program and provides adequate background information on the theory behind the specific computations, a description of formulas and methods, and examples with output samples. The chapters are alphabetically arranged by utility name (which has no relation to the type of calculations performed). For instance, ATFRAC, CASECONT, LOGISTIK, MANTELX and MATCHED provide computations for the analysis of case-control studies, but their chapters are not arranged within a single section. That said, I do not find this to be a problem. Most readers will probably refer to the book as guidance for using a specific program and will thus welcome the ease with which the chapters are laid out by software name. An extensive reference list and an index with a program finder are valuable additions. The PEPI collection is available for free (http://www.sagebrushpress.com/pepibook.html)–access that is much appreciated by colleagues in resource-poor regions. It cannot replace the commercial software packages for in-depth analysis of large datasets, mostly because of the lack of extensive regression procedures and data editing and management tools. Regardless, “Have PEPI, will travel” is the motto here. If given only one choice for a program that epidemiologists would want to have in their portable PCs, PEPI would be the one.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.