This paper presents a novel approach to determine the sensitization probability of a non-robustly testable path using probability density functions (PDFs). The proposed approach systematically refines a set of patterns that sensitize the path non-robustly which initial set has been derived with existing methods, and is kept implicitly. Accurate measure of the sensitization probability is obtained fast by avoiding Monte-Carlo. It is shown experimentally that the proposed approach is accurate and much faster than Monte-Carlo, and thus can be used to rank a collection of non-robust paths considering their sensitization characteristics.