The statistical distributions of fatigue life data of stampable CF/PA composite at different stress ratios are identified, and a procedure for predicting the S-N relationships for any stress ratios that consider the influence of probability of failure P is proposed. Constant amplitude fatigue tests are first performed at different stress ratios R = 0.1, 10, χ (≡ σ_UCS/σ_UTS), respectively. To identify the statistical nature of fatigue life, a lognormal distribution and a Weibull distribution are applied to the fatigue data obtained, respectively. The anisomorphic constant fatigue life (CFL) diagrams for different probabilities of failure (P = 10%, 50% and 90%) are constructed using the P-S-N curves for the critical stress ratio. It is shown that the P-S-N curves for different stress ratios can successfully be predicted using the probabilistic methodology proposed in this study.