A systematic method has been developed for the construction of neutronic data bases used with the GRASS code in the analyses of postulated transients in Savannah River Plant (SRP) reactors. In this context, a neutronic data base consists of a set or sets of correlation equations that represent the few-group macroscopic cross sections for various cell types in the reactor charge. The cross sections are functions of the reactor state variables (temperatures, densities, material compositions, etc.). To minimize the size of the data set library required to support the correlations, two steps are taken. First, the state variables are grouped into separate correlations, or “regimes.” A state variable is assigned to a particular regime according to the relative magnitude of the reactivity perturbation caused by changing the value of the variable over its allowed range. The separate regime correlation equations are combined linearly within GRASS to produce a single set of few-group cross sections for each cell type at any time within the transient. The second step involves the use of Box-Behnken or other incomplete factorial data library designs to support each regime correlation. The derivation of the regime correlation equations is performed with the CRASS system of codes. The above methods are demonstrated by application to the design of a transient data base for a typical SRP reactor charge. Transient phenomena included are the temperature feedback of normal reactor operation, coolant flow instability, assembly melting with associated entrainment of melted particles in the moderator, moderator boiling, and gadolinium poison injection.