be able to proceed with his or her study as initially planned. Finally, principal components analysis works best with a large number of observations, which may not be possible in many investigations ( Short & Horn, 1984).
Each method has its advantages and disadvantages that must be weighed by an investigator according to the parameters of a particular study. Many of the methods discussed previously require calculations (e.g., means, standard deviations, correlations) after the data have been collected. This may have serious implications for the researcher and preclude the use of response-guided experimentation. Response-guided experimentation involves inspection of data as they are collected in order to allow the data to guide the experiment. Some researchers espouse this procedure ( Sidman, 1960), whereas others find it less favorable (e.g., Allison, Franklin, & Heshka, 1992; Edgington, 1984). For example, Allison et al. ( 1992) reported that "When visual inspection is combined with response guided experimentation, Type 1 error rates could easily be high as 25%" (p. 45).
The formation of a behavior composite is conditional upon two criteria. First, the component variables should correlate substantially. Second, the component variables should have close to equal variances and covariances. When these two criteria are met, investigators can be reasonably confident that a behavior composite should be formed. When these two criteria are not met, investigators should be inclined to separately analyze each component variable.
Science and practice are both at their best when built on a strong foundation of measurement. We can only communicate with each other about "such-and-such" when we can say exactly what we mean by "such-and- such." One of the best ways to do this is to specify how we measure "such-and-such." Measurement provides precision to our ideas, theories, and constructs, and we hope that this chapter assists investigators in providing precision in their measurements.
Allen M. J., & Yen W. M. ( 1979). Introduction to measurement theory. Monterey, CA: Brooks/Cole.
Allison D. B., Franklin R., & Heshka S. ( 1992). "Reflections on visual inspection, response guided experimentation, and Type 1 error rate in single-case designs". Journal of Experimental Education, 61, 45-51.