The conventional practice for testing analog or RF integrated circuits is specification-based testing, which relies on the direct measurement of the circuit performance parameters. This approach offers good test quality but at the price of extremely high testing costs. In order to reduce test costs, a promising approach, called indirect or alternate testing has been proposed. Its basic principle consists in using the correlation between the conventional analog/RF performances and some low-cost measurements, called Indirect Measurements (IMs), in order to estimate the analog/RF parameters without measuring them directly. In this paper, we perform efficiency evaluation of this strategy, and in particular we perform a comparative analysis of different IM selection strategies in order to define efficient alternate testing implementation. Efficiency is evaluated in terms of model accuracy by using classical metrics such as average and maximal prediction errors, and in terms of prediction reliability by introducing a new metric called Failing Prediction Rate (FPR). Results are illustrated on two case studies for which we have experimental test data.