Data mining methodology is used to study the content of presidential State of the Union messages. Ratings of greatness in leadership from C-Span panels of historians are defined as the dependent variables. The set of independent variables included measures of values, rhetorical style and intellective ability. Results of linear regression models with bootstrapped standard errors indicated a value facet related to self-direction and stylisticvariables to significantly discriminate the expert ratings of presidential leadership by historians.
Steven D. Silver. "Discriminating Rated Leadership in the U.S Presidency from Value Content and Stylistic Variables in State of the Union Addresses: A Data Mining Methodology" 47th Decision Sciences Institute Annual Meeting (2016).