Herman Rubin wrote: > In article <[EMAIL PROTECTED]>, R. Martin <[EMAIL PROTECTED]> wrote: > >>Art Kendall wrote: > > >>><part 2> >>>If you have re-entered the data, or re-run the experiment, and done very >>>thorough exploration of the data, you are stuck as a last resort with >>>doing multiple analyses: including vs excluding the case(s); changing >>>the values for the case(s) to hotdeck values, to some central tendency >>>value, or to max or min on the response scale (e.g., for achievement, >>>personality, or attitude measures), modeling the specialness of the >>>particular value, etc. > > >>>A very good book on regression is: >>>Cohen, Jacob, et al (2003) Applied multiple regression/correlation >>>analysis for the behavioral sciences, third edition. Mahwah, NJ. >>>ISBN 0-8058-2223-2 >>>LoC HA31.3 .A67.2003 >>>Outliers are discussed though out the book. > > > I have not seen this, but I am greatly suspicious.
The earlier editions of the Cohen & Cohen book describe how to add a binary variable that codes for whether or not a person/subject has valid data on a given predictor variable. Dave Howell's page has a brief summary and an example: http://www.uvm.edu/~dhowell/StatPages/More_Stuff/Missing_Data/Missing.html > > BTW, the earliest treatment I know for outliers was for > astronomy and physics in the 19th century. There, the > effect of improperly discarding suspected outliers was > primarily an affordable loss of efficiency, but biases also > resulted. The published values of the speed of light kept > decreasing until really good observations were made, as the > first values were high, and outliers were rejected, partly > using previous experiments. > > >>>The best way to deal with outliers is to prevent them through thorough >>>quality assurance efforts in the data gathering (measurement), data >>>entry, and exploratory analysis phases of the research. > > >>But that takes thinking and work, and statistics is supposed to save >>us all that effort, right? ;-) > > > Anyone who takes this attitude should not try to do > anything with data. The results are very likely to be BAD, > and this IS the case with much of the statistical studies > in psychology, education, and medicine. > > Until one understands decision problems, it would be wise > NOT to take any methods courses. I do believe that Art Kendall had his tongue firmly in his cheek when he made that comment about thinking and work. Cheers, Bruce -- Bruce Weaver [EMAIL PROTECTED] www.angelfire.com/wv/bwhomedir/ . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
