Statistically significant sites

Most of us know less about statistics than we feel we should but, even as a scientist, it’s often not necessary to be an expert. When it comes to presenting your data, or reviewing someone else’s, however, it can help to have a ready reference at your fingertips. Before you run out and buy a textbook or ask advice from a colleague, take a look at what the web has to offer. Sites offering online stats textbooks provide easy-to-locate, nontechnical statistical information for researchers. A New View of Statistics, an extensive textbook by Will Hopkins of the University of Otago in New Zealand, is probably the easiest to read of these online resources. Hopkins uses a conversational writing style, not equations, to describe concepts. Hopkins has chosen most of his examples from the field of exercise physiology, but this doesn’t limit the usefulness of this resource for scientists from other fields. Hopkins starts with the very basics, including Summarizing Data: Simple Statistics and Effect Statistics and moves on to discuss more complex statistics, including multiple linear regression. The site also includes discussions of sample size estimation and other subjects of interest to researchers. His topics pull you in because some of them have the air of opinion pieces, including Mean ± SD or mean ± SEM?, which discusses which one you should publish. The Electronic Statistics Textbook from StatSoft lies at the opposite end of the spectrum, stylistically speaking. This web site offers plenty of useful information, written in a technical and sometimes dry manner, for both statistically naive and experienced scientists. The site might have a special appeal for scientists in specific fields, because it covers some rather rare statistical topics, including neural networks and cluster analysis. Selecting the latter topic lets the reader learn about the general nature of cluster analysis or read a more detailed description covering linkage rules and different algorithms used in these analyses. The site also features a Statistical Advisor. This online decision tree is a good reference for those without a local stats expert. A researcher might have a whole raft of statistics tests on a computer, but which test is the right one? The Statistical Advisor leads visitors through a series of questions. Each layer of questions grows more specific, ultimately leading to the chapters in the textbook that discuss the relevant tests. Visitors need some basic understanding of statistics before beginning a session with the advisor, but the descriptions are written clearly and refer back to specific sections in the online text. One more online textbook worth mentioning is HyperStat Online , which was developed by David Lane of Rice University. This site is one component of the Rice Virtual Lab in Statistics. This introductory-level text is especially well organized and easy to navigate. Two other useful sites for researchers are Pitfalls of Data Analysis and Intuitive Biostatistics: Choosing a Statistical Test. Both of these easy-to-read pages deal, in a relatively short fashion, with more specific topics. The former was written by Clay Helberg and the latter is a chapter from a book by Harvey Motulsky. After you’ve perused your favorite online stats text, Interactive Statistical Calculation Pages can get you started with the actual data analysis. This site claims to have more than 300 links to web pages that perform statistical calculations. John Pezzullo of Georgetown University did a nice job of organizing and annotating these sites. He also noted which sites use JavaScript or Java applets. Pezzullo’s own Probability Distribution Functions serves as a quick and easy way to find exact p-values or to look up a critical value for a given p-value. From one page you can find the normal, Student’s t, chi-square or Fisher F distribution. Sometimes, being knowledgeable is more a matter of knowing where to look things up than actually knowing them off the top of your head. With a few stats sites bookmarked in your web browser, you’ll be able to hold your own when the conversation next turns to p-values and Poisson distributions.