On testing common indices for two multi-index models: A link-free approach

We propose a link-free procedure for testing whether two multi-index models share identical indices via the sufficient dimension reduction approach. Test statistics are developed based upon three different sufficient dimension reduction methods: (i) sliced inverse regression, (ii) sliced average variance estimation and (iii) directional regression. The asymptotic null distributions of our test statistics are derived. Monte Carlo studies are performed to investigate the efficacy of our proposed methods. A real-world application is also considered.