Robert J. MacG. Dawson wrote: > > Thom Baguley wrote: > > > This allows us to generalize to a hypothetical population of > > similar data sets > > This is a tricky statement, with more to it than meets the eye. On the > one hand, we do usually generalize to a population of *individuals* > rather than of *sets*. On the other hand, the observation generalizes > not to the individuals in the real world, but to observations of them - > data. And there are some types of observation effect (learning effects, > fatigue, etc) that are evident only in the set, and cannot be observed > in one datum.
I'm not happy with the data set phrasing - I actually think that we generalize about processes that generate individual data points (where "process" is defined liberally and includes individual people). > (it is a misconception to think that a specific population - > > such all students or all students in Kentucky, or all students wearing blazers > > - is being generalized to). > > Again, there is a lot in this, but perhaps it would be more accurate to > say that the process of generalizing to a wider population is > extrastatistical, and separate from the statistical process of Almost 100% in agreement - statistical generalization is only a tiny bit of the picture. I don't think it is _entirely_ separate however - typically one is a precursor to the other. If you didn't have a case for statistical generalization I think there would rarely be a point in engaging in extrastatistical generalization. > generalizing to which Thom alludes. If we really believed Thom's words > literally, the literature would be full of papers entitled "Effects of > Fatigue on the Reaction Times of Students in the Psych 100 Subject Pool > at Euphoric State U. (Class 0f '05) On Sept. 30, 2003" and, given its > irrelevance to anybody else, nobody would sponsor such research except > the subjects' parents. Even physics journals would have papers like > "Abrasion by Airborne Particles of Pieces From One Slab of Styrene That > My Lab Assistant Lost Last Week So, No, You Can't See It." Experimental > science would end up, like Borges' character "Funes the Memorious", > bewildered by a cloud of unrelated minutiae. That's more-or-less the opposite of what I was trying to say. The population your are generalizing in a statistical sense is defined by how the orginal data points were sampled (i.e., the process the generated those data points). My point was that in standard inferential statistics you can't point to a specific population of people (or whatever) that is being generalized to. Simply by using inferential statistics you are generalizing to other situations - though not necessarily the situations you want to generalize your conclusions to. > To make any use of these results, we have to decide what populations > they *do* apply to. We have to realize that it's *not* necessarily what > we'd like; we have to realize that the stats won't do it for us, though Again, 100% in agreement. You need to understand the context in which the data were collected, the relevant theory and so forth in order to make such generalizations. > common sense may; and we have to accept that the bigger the population > the less the application. I'm not sure what you are referring to here - I can think of at least two different interpretations. Thom . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
