<![CDATA[This study aims to evaluate the performance of group-specific resampling in a truncated spline-based semiparametric multigroup path analysis model with heterogeneous relationship structures across groups. The study used a simulation approach based on empirical patterns in AI Literacy data. The multigroup concept is represented by the interaction of dummy variables within a single semiparametric model, thereby allowing for differences in the structures of linear and nonlinear relationships across data groups. The evaluation was conducted across four resampling method combinations: bootstrap-bootstrap, jackknife-jackknife, bootstrap-jackknife, and jackknife-bootstrap. Method performance was evaluated using the average bias, average standard error, standard error ratio, and coefficient of determination ( ). The results indicate that all method combinations yield relatively consistent average bias values. However, the jackknife-jackknife combination produce the smallest average standard error and standard error ratio, thereby providing better inference stability compared to the other combinations. The results also indicate that the jackknife method tends to be more adaptive for nonlinear relationships based on truncated splines. Thus, the selection of resampling methods for semiparametric multigroup models must consider the characteristics of the relationships within each data group to yield more stable and accurate inferences.]]>
Copyrights © 2026
