The research activity scale includes the following correlates of research activity: research & development (R&D) expenditures in science and engineering; R&D expenditures in non-S&E fields; S&E research staff (postdoctoral appointees and other non-faculty research staff with doctorates); doctoral conferrals in humanities fields, in social science fields, in STEM (science, technology, engineering, and mathematics) fields, and in other fields (e.g., business, education, public policy, social work). These data were statistically combined using principal components analysis to create two indices of research activity reflecting the total variation across these measures (based on the first principal component in each analysis).
One index represents the aggregate level of research activity,
and the other captures per-capita research activity using the expenditure and staffing measures divided by the number of full-time faculty within the assistant, associate and full professor ranks. The values on each index were then used to locate each institution on a two-dimensional graph. We calculated each institution's distance from a common reference point, and then used the results to assign institutions to one of three groups based on their distance from the reference point. Thus the aggregate and per-capita indices were considered equally, such that institutions that were very high on either index were assigned to the "highest research activity" group, while institutions that were high on at least one (but very high on neither) were assigned to the "higher research activity" group. Remaining institutions and those not represented in the NSF data collections were assigned to the "limited research activity" category. Before conducting the analysis, raw data were converted to rank scores to reduce the influence of outliers and to improve discrimination at the lower end of the distributions where many institutions were clustered. Detailed information about how the research activity index was calculated can be found here.
http://carnegieclassifications.iu.ed...ity_Index.xlsx