Try this: When a measurement in group A is linked specifically to a measurement in group B, we can discuss paired measurements.
Examples: Two different paints are compared, by painting them on the same substrate sample, half of each substrate for each paint. Now, differences in paint performance due to substrate prep will cancel out, and we can more easily see differences due to paint source. Differences in opinion due to female and male sources: We query married couples seperately, and get their opinions on lots of subjects, using a Likert scale (!). Then we see if the differences between gender, _within_ a married couple, are much different. Does one test sample (such as a TB test) produce a larger response than a different type of test material? We apply test 1 to one side of the patients, and test 2 to the other side. Do we see differences in response, _on the same patient_? In each case there is a linkage, a commonality, for each measurement in the two groups. By using a paired t test, we reduce the variance, and have a mosre sensitive test. Cheers, Jay > [EMAIL PROTECTED] (Donald Burrill) wrote: > > >... nor why you > >have asked the question. > > I have asked the question because I haven't grasped the concept of > paired vs unpaired data completely yet. The statistics books I read so > far only gave specific examples for pre-/post-test data and only > mentioned that there can also be other paired data such as from > daughter/mother pairs, couples, twins, siblings, etc. > > First, I thought the subjects yielding paired data have to homogeneous > or as similar as possible, a condition which is perfectly fullfilled > for repeated measurements on the same subjects, but since siblings and > couples aren't necessarily very alike this can't really be a > criterion. > > The only required criterion seems to be that each measured variable > can be assigned unambiguously to another one, which is true for all > the above cases. > > Rich Ulrich gave male/female as an example but I can't see how I can > unambiguously assign one male to one female for comparing the measured > variables. > > Peter > . > . > =================================================== ============== > Instructions for joining and leaving this list, remarks about the > problem of INAPPROPRIATE MESSAGES, and archives are available at: > . http://jse.stat.ncsu.edu/ . > =================================================== ============== > -- Warner Consulting, Inc. 4444 N. Green Bay Road Racine, WI 53404-1216 Ph: (262) 634-9100 email: [EMAIL PROTECTED] Home of the A2Q Method(tm) What do you want to improve, Today! via CoreComm Webmail. http://home.core.com . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
