This research sought to determine whether any measure(s) used in the Carnegie Foundation's classification of Doctoral/Research Universities contribute to a greater degree than other measures to final rank placement. Multilevel Modeling (MLM) was applied to all eight of the Carnegie Foundation's predictor measures using final rank (doctoral/research, high research, very high research) as the outcome (dependent) variable. Data came directly from the Carnegie Foundation. One additional variable, private or public control came from the IPEDS Peer Analysis System. All measures used in the Carnegie Foundation's analyses exhibited strong interrelationships (multicollinearity), which reduces the reliability of multivariate analyses. The overall MLM regression model predicted approximately 50% of the variance in rank, with an estimated multiple r of 0.72. The most powerful predictor of rank was federal science & engineering (S&E) expenditures. Once this variable entered the prediction model, only doctorates granted in the humanities added significantly to prediction (3.5% of variance). Due to the multiplicity inherent to MLM analyses, significance for all tests was set at p less than 0.001. Although both the number of post doctoral appointments (Spearman Rranks = 0.86) and non-faculty researchers (0.67) exhibit strong simple relationship with rank when S&E expenditures and humanities doctorates are entered into the MLM model, the unique contribution of both post doctoral appointments and non-faculty researchers proved to add both non-significant and negative increments in predicting Carnegie rank. Most simple relationships between predictor variables and the outcome Carnegie rank ranged between 0.75 (number of faculty) and 0.89 (S&E expenditures). All of the measures also exhibit strong relationships with other predictors. That the number of faculty has a simple Rranks of 0.75 indicates that a research institution's size alone relates to their rank. Using eight predictor measures, all interrelate strongly and significantly, is effectively like using a single measure to rank institutions. An institution's S&E expenditures may thus be effectively used as that single predictor, although doctorates in the humanities can also influence an institution's rank. Appended is: Reasons for Using Multilevel Modeling Rather than OLS Statistics. (Contains 3 footnotes, 1 figure, and 3 tables.) [This report represents an Internal Technical Report, Office of Planning and Analysis, University of South Florida, Tampa, Florida]