Abstract
This study proposes a semiparametric path analysis framework that integrates truncated spline modeling with resampling-based inference to capture both linear and nonlinear relationships within a unified structure. The motivation arises from the limitation of conventional path analysis, which relies on linearity assumptions that are often violated in empirical data, as indicated by the Ramsey RESET test. To address this issue, a truncated spline approach is employed to flexibly model nonlinear relationships, while statistical inference is conducted using double resampling techniques. A simulation study is performed to evaluate the performance of resampling methods under varying conditions. The results show that for a sample size of n=200 with a single nonlinear relationship, the double jackknife method provides more stable and efficient estimates compared to alternative approaches. This finding motivates its application in the empirical analysis. The empirical results, based on data from East Java, Indonesia, reveal that technology access has a significant direct effect on both financial knowledge and financial literacy. A nonlinear relationship is identified between technology access and financial literacy, characterized by a threshold effect captured through truncated spline modeling. However, the indirect effect through financial knowledge is found to be statistically insignificant. Overall, the proposed approach offers a flexible and robust framework for modeling complex causal relationships and improves inference accuracy in semiparametric path analysis.