Currently available group-contribution methods for T,, P,, and V, were evaluated using the Ambrose (1980) critical property data compilation. The Ambrose estimation methods were found to be the most accurate. Linear regression methods were also employed to develop alternate estimation methods which were found to have an accuracy comparable to those of Ambrose.K. M. KLINCEWICZ SCOPECritical properties (T,, P,, V, ) are often required when using generalized property estimation methods. Due to the paucity of experimental values, one must often estimate these properties, and a group-contribution approach is generally used. The method developed by Lydersen (1955) is the most widely employed. However, many new experimental critical properties have become available since the Lydersen report and several alternate estimation schemes have been proposed. In the first portion of this paper, a few methods have been briefly reviewed and evaluated for accuracy using the currently available data.With the rapid advances in linear regression programming, it was felt to be of value to use this methodology to optimize the coefficients and group contributions utilizing a number of objective functions involving T,, P,, or V,. In this manner, one can rapidly test a number of objective functions, selected group-contribution substructures, auxiliary variables, etc. to obtain the least error method of estimation. CONCLUSIONS AND SIGNIFICANCEEvaluating all available estimation methods for T,, P,, and V, , it was shown that the methods of Ambrose (1979Ambrose ( , 1980b were the most accurate. These methods employ Tb/( T, -Tb), (MW/Pc)l12, and V, as objective functions (Q) with the equation forms given in Table 1 and the necessary group contributions in Table 3. The statistics relating to expected errors are shown in Table 5.Linear regression techniques also provided group estimation methods close in accuracy to those of Ambrose but with a simpler set of structural subgroups. In this type of analysis, multicollinearity was identified as a serious problem and the final choice of subgroups minimized the difficulties encountered.The new proposed estimation equations are given as Eqs. 10 through 12 with group contributions in Table 6. Expected errors are shown in Table 7.Models not involving the concept of group contributions were able to explain a large portion of the variance in the critical property data. Reasonably accurate estimation methods could be found where T,, (MW/P,)112, and V, were correlated only with Tb, molecular weight and the number of atoms in the compound. These preliminary results suggest that factor analysis may provide a viable alternate route to group-contributions in the estimation of critical properties.The properties of one material are often related to properties of another by assuming the respective intermolecular potentials of both are similar if the energy and separation variables are scaled appropriately. Materials that behave in this manner are called conformal substances (Rowlinson, 1969). Utilizing relation...