Dataset for Non-linear Modeling of Behavioral Realism and Perceived Agency in Virtual Influencer Engagement

Overview: This dataset contains primary survey responses from 720 Generation Z Instagram users, collected to investigate the non-linear effects of Behavioral Realism (BR) and Perceived Agency (PA) on User Engagement (UE) with AI-generated virtual influencers (VIs). This repository provides a complete replication package, including the raw data, the unstandardized results from the Multivariate Adaptive Regression Splines (MARS) analysis, and the R script used for non-linear estimation. File Specifications: The repository consists of three files necessary for full computational reproducibility: a) Data.xlsx: The anonymized master dataset containing all survey items and constructs. b) MARS Unstd.xlsx: The unstandardized output file of the latent variable scores extracted from SmartPLS c) Non-Linearity using MARS Algorithm.R: The custom R script used for the secondary analysis.