Measurement Scales (4 of 6)

---

The first is philosophical and challenges the validity of the notion that there is some unseen "true" measurement scale that is only being approximated by the rating scale. The second counter argument accepts the notion of an underlying scale but considers the examples to be very contrived and unlikely to occur in real data. Measurement scales used in behavioral research are invariably somewhere between ordinal and interval scales. In the preference experiment, it may not be the case that the difference between the ratings one and two is exactly the same as the difference between five and six, but it is unlikely to be many times larger either. The scale is roughly interval and it is exceedingly unlikely that the means on this scale would favor color displays while the means on the "true" scale would favor the B & W displays.

There are some cases where one can validly argue that the use of an ordinal instead of a ratio scale seriously distorts the conclusions. Consider an experiment designed to determine whether 5-year old children are more distractible than 10-year old children.