Enter your email address to receive instant alerts when replies are posted to the messages below. If you choose at any time to stop receiving alerts, simply click on the "unsubscribe" link at the bottom of any email alert you receive.
You raise a good question. The short answer is that there is no completely agreed upon way to handle outliers, statistically speaking.
My general recommendation when working with data is to retain all outliers unless 1) very few outliers cause the data to violate major assumptions of parametric statistical procedures, 2) the few outliers are clearly beyond +/- 2SD of the mean within a single measurement item's category, and 3) the removal of such outliers does not drastically alter the message provided by the descriptive data.
Now, if the outliers might be meaningful, there are many ways to work with them. For instance, you could use non-parametric procedures.
I hope my response helps some; unfortunately, data analysis is not something entirely concrete and rigid.
- Jay L. Michaels Assistant Professor of Psychology Presbyterian College 503 S. Broad Street Clinton, SC 29325
Forum Home Page
