What's on the web? Statistics for medical students

Mention statistics to the average medical student and chances are the response will be boredom tinged with trepidation. Yet, as doctors, we will encounter data analysed and presented in many different forms and must somehow make sense of it all.

The internet contains lots of statistical resources, including several complete textbooks. However, most of these are almost painful to read: "When the word 'mean' is used without a modifier, it can be assumed that it refers to the arithmetic mean....the formula in summation notation is µ=Sx/N," says the Hyperstat Online Textbook (http://davidmlane.com/hyperstat), a comprehensive statistics text and more, which is, admittedly, not aimed at medical students. Each chapter links to other online texts, statistical instruments, and even amazon.com, where you can buy your very own copy of the recommended further reading.

STATS (www.childrens‐mercy.org/stats/index.asp), or Steve's Attempt to Teach Statistics, is slightly easier to understand. Steve Simon, a statistics consultant at Children's Mercy Hospital, has produced a collection of handouts. Simon uses data from medical journal in his worked examples, and his informal writing style somehow succeeds in making statistics both readable and interesting. He explains, "The mean is simply the average of all the items in a sample.... add up all the sample values and divide by the size of the sample." Everything you need to know and not a symbol in sight. The only thing STATS is short on is graphics, though that is probably also why it loads so quickly. Statsoft (http://statsoft.com/textbook/stathome.html) on the other hand, is full of them, from animated bell curves to disappearing outliers-the works. It even has its own search engine.

Having got to grips with the relevant statistical concepts, the next step was to choose a test for my own data. Statistics at Square One (http://bmj.com/collections/statsbk/index.shtml) is a textbook available online. The chapter called "Study design and choosing a statistical test" has a handy table that matches data to test. A more interactive way of doing this is provided at the home of the statistical test finder (http://ase.tufts.edu/psychology/stat.htm). Enter the characteristics of your sample, and it takes you through the steps of choosing the right test.

For actual analysis of data, I use SPSS, for which there is a good introductory tutorial at www.shef.ac.uk/~scharr/spss/index2.htm However, SPSS is just one of many statistical software package. http://members.aol.com/johnp71/javastat.html contains over 600 regularly updated links to free or "free but" statistics programs, such as Webstat (www.webstatsoftware.com) and a demo version of Minitab (www.minitab.com/products/13/demo/index.htm). If all else fails, try the departmental statistician; www‐phm.umds.ac.uk/Statinfo/hints.htm gives hints to help you get the best out of one.