On Tue, 12 Feb 2002 12:27:08 -0600 "Paul C. Smith" <[EMAIL PROTECTED]> wrote:
> But I assume that also means > that they're not likely to encounter much in the way of grouped data. Sure, > I can think of some situations in which they might (secondary analyses of > previously grouped data), but I've noticed that those things seem to make > students thing that it's _desirable_ to group data, even when there is no > reason to do so. When they design surveys, students tend to want to ask age > or number of children (for example) using ranges (which is just simply > wrong, in my opinion, given that people can easily tell you their exact age > or number of children, so you get good data). When they code good data > (exact number of children, for example), they STILL want to come up with > ranges (in short, to simply throw away good data, by created grouped data). > I'm in favor of doing whatever I can to nip that whole thing in the bud, and > ignoring the whole grouped data thing seems like a good place to start. > > (let the stoning begin... <grin>) > I agree with much of what John, Paul, and Miguel have said with one very big exception.... In an important sense, all data is grouped-data. We specify some boundary conditions of inclusion/exclusion in a unit and those conditions typically cover a range of variations which we *might* attend to under other circumstances. To continue with and (likely) mangle Paul's example, investigators may be making a mistake by creating intervals of 1-5 people, 6-10 people. However, we are still dealing with grouped-data when we count "people." We may define a person as a living homo sapiens. This groups males and females into the count. If we split our measure into males and females, then we may be ignoring information about age, race, ice cream flavor preference and so on. And these factors are "grouped" in our count. Our measurement units are functional groups, the sizes of which are best defined by the questions of interest. We must always deal with the problem of grouped-data. (Now let the stones really fly :-) Ken ---------------------- Kenneth M. Steele, Ph.D. [EMAIL PROTECTED] Dept. of Psychology Appalachian State University Boone, NC 28608 USA --- You are currently subscribed to tips as: [EMAIL PROTECTED] To unsubscribe send a blank email to [EMAIL PROTECTED]
