Using simulated redshift surveys generated from a high resolution N-body cosmological structure simulation, we study algorithms used to identify groups of galaxies in redshift space. Two algorithms are investigated; both are friends-of-friends schemes with variable linking lengths in the radial and transverse dimensions. The chief difference between the algorithms is in the redshift linking length. The algorithm proposed by Huchra \& Geller (1982) uses a generous linking length designed to find ``fingers of god'' while that of Nolthenius \& White (1987) uses a smaller linking length to minimize contamination by projection. We find that neither of the algorithms studied is intrinsically superior to the other; rather, the ideal algorithm as well as the ideal algorithm parameters depend on the purpose for which groups are to be studied. The Huchra/Geller algorithm misses few real groups, at the cost of including some spurious groups and members, while the Nolthenius/White algorithm misses high velocity dispersion groups and members but is less likely to include interlopers in its group assignments. In a companion paper we investigate the accuracy of virial mass estimates and clustering properties of groups identified using these algorithms.

Two algorithms for the identification of galaxy groups from redshift surveys are tested by application to simulated data derived from N-body simulation. The accuracy of the membership assignments by these algorithms is studied in a companion to this paper (Frederic 1994). Here we evaluate the accuracy of group mass estimates and the group-group correlation function. We find a strong bias to low values in the virial mass estimates of groups identified using the algorithm of Nolthenius \& White (1987). The Huchra \& Geller (1982) algorithm gives virial mass estimates which are correct on average. These two algorithms result in group catalogs with similar two-point correlations. We find that groups in a CDM model have excessively large mass to light ratios even when the group richness distribution agrees with observations. We also find that our CDM groups are more strongly correlated than individual halos (galaxies), unlike the groups in the CfA redshift survey extension.