Manifestation of hormesis in longevity was modelled by modification of the mortality rate during and after the period of a stress factor action. In heterogeneous population this can lead to observation of unchanged mortality during action of the stress and decrease in mortality after stress period. Stochastic simulations were made to investigate the possibility of detecting the hormesis effect on the basis of the stress-control longitudinal data. The goal of the stochastic simulation was to investigate the role in the hormesis detection of control and stressed group size, of population heterogeneity variance value, of stress and hormesis attributable risks as well as the role of a prior information about the survival in the control group. It was demonstrated that if the attributable risks for stress and hormesis effects are approximately equal, then in both 'high' and 'low' heterogeneous populations the hormesis phenomenon is detected with probability higher than 75% even in relatively 'small' groups of 50 subjects. In case of 'weak' effect the hormesis phenomenon is not detected in a 'highly heterogeneous' population even in a group composed of 1000 subjects. In a 'low heterogeneous' population the hormesis phenomenon is detected with probability higher than 70% when the group size is not less than 200 subjects. Information about the survival in control group did not play a critical role in all experiments and exact survival curve may be replaced by the traditional Kaplan-Meier estimate.