In article <[EMAIL PROTECTED]>, [EMAIL PROTECTED] writes >I am looking for a statistics book that would cover some in-depth >statistical analysis that could be handy for laymen. > >I enjoyed my college statistics course as part of my computer >information systems degree, and I would like to learn additional >techniques that could be used in everyday situations (data quality >analysis, fantasy football, consumer budgeting). > >Can anyone recommend a good book? Have you read "The Elements of Graphing Data" and "Visualizing Data" by W.S. Cleveland? They are a good read, and since they show you how to display data in the most informative way, they should fit in well with your computer systems experience.
They concentrate much more on showing patterns in data than in distinguishing those patterns from random fluctuations. I suspect that for most everyday decisions, your background knowledge is so strong and the amount of data available so small that traditional statistical inference may not actually pay off. If (like me) you are interested in it as an amateur for its own sake, you might be as well to make sure that you understand the basics by applying them in practice. I have been trying to run simple experiments for fun (and am learning about birdwatching in support of this). So far I have found that where the answer is uncertain enough for me to be interested in the outcome, I don't get a definitive answer from statistics in the amount of data I can collect before I lose interest. I am now strongly motivated to look for ideas to reduce noise and increase statistical power! (I should probably also look for experiments that give me more than one data point per outing). If you do want to get into inferential statistics, try "Practical Nonparametric Statistics", by Conover. Examples of this that you might have covered already are the sign test and the Mann-Whitney/Wilcoxon test. The preface says that this is intended both as a one-semester course in this area of statistics and as a quick reference to the most useful parts of the subject for research workers using statistics. I like this sort of stuff because you can understand the test as a whole, and not just apply a recipe cooked up by somebody else, and because you can often get exact results even for very small data sizes. Of course, that exact result might be that your data don't really say anything one way or another. -- A. G. McDowell . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
