Ade Asmi
Department of Civil Engineering, Bakrie University Indonesia

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

LATENT HETEROGENEITY IN DECISION MECHANISMS FOR FUTURE ELECTRIC VEHICLE PURCHASE AMONG CURRENT OWNERS: A FIMIX-PLS ANALYSIS Fredy Jhon Philip Sitorus; Leksmono Suryo Putranto; Ade Asmi
Jurnal Pensil : Pendidikan Teknik Sipil Vol. 15 No. 3 (2026): Jurnal Pensil : Pendidikan Teknik Sipil
Publisher : LPPM Universitas Negeri Jakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21009/jpensil.v15i3.67703

Abstract

Future electric vehicle (EV) purchase intention among current owners may be formed through different decision mechanisms even when an aggregate structural model produces stable average coefficients. This study examines unobserved heterogeneity in a survey of 521 current EV owners in Jabodetabek, Indonesia. The aggregate PLS-SEM model is retained only as the analytical baseline, whereas finite mixture partial least squares (FIMIX-PLS) is used to compare latent structural solutions with two, three, and four segments. The three-segment solution was retained because BIC (2677.79) and CAIC (2739.79) reached their lowest values at three segments, and the estimated mixture proportions (53.0%, 37.8%, and 9.2%) were less imbalanced than the two-segment solution (89.9% and 10.1%). AIC3 favoured four segments, whereas entropy and NFI favoured two; the retained solution is therefore a model-selection compromise rather than a uniquely determined partition. In the two larger segments, perceived behavioural control had the largest proximal coefficient in Segment 1 (β = 0.497), whereas subjective norm had the largest coefficient in Segment 2 (β = 0.468). The smallest segment had near-perfect explained variance for purchase intention (R² = 0.9998) and a residual variance of 0.001, so its coefficients are interpreted cautiously. The analysis adds segment-specific information to the aggregate model but does not establish statistically significant differences between segment coefficients.