Integrating Non-metric Data in Partial Least Squares Path Models: Methods and Application

Bioregions are predominantly climatic for fishes of northern lakes

“…PLS methods can also be performed with (non-metric) nominal or ordinal indicators by introducing an optimal scaling step (Petrarca et al, 2017). Here, non-metric variables are assigned values on an interval scale by optimizing additional scaling parameters as part of the iterative PLS-SEM process.…”

“…To evaluate the roles of spatial and environmental constraints on fish communities, we applied non‐metric partial least‐squares structural equation modelling (NM‐PLS‐SEM, also called PLS path modelling) paired with multiple logistic regression (Petrarca et al., 2017). PLS‐SEM is a variance‐based approach to estimating unobserved latent variables and their relationships by a series of simple or multiple ordinary least‐squares (OLS) regressions (Sanchez, 2013).…”

“…Thus, the algorithm offers an intuitive inferential framework while also applying dimension reduction to develop proxies for multidimensional concepts, such as connectivity, climate and habitat (e.g., Loewen et al., 2020). PLS methods can also be performed with (non‐metric) nominal or ordinal indicators by introducing an optimal scaling step (Petrarca et al., 2017). Here, non‐metric variables are assigned values on an interval scale by optimizing additional scaling parameters as part of the iterative PLS‐SEM process.…”

“…The approach was also applied to multilevel categorical outcomes of our modularity analysis. However, given that estimating a binomial response by OLS is less than ideal, even with the optimal scaling algorithm, we used a modified approach (Petrarca et al., 2017), initially obtaining latent variable scores from NM‐PLS‐SEM and then re‐estimating inner model path coefficients by multiple logistic regression (generalized linear regression with binomial error distribution and a logit‐link function; glm function in base R). The explanatory power of logistic models was measured by Tjur’s R 2 using the RsqGLM function in the “modEvA” package (Barbosa et al., 2015).…”

Pengaruh Ketangkasan Rantai Pasok Terhadap Kinerja Bisnis Kedai Kopi Di Kota Malang

“…Tahap selanjutnya adalah melakukan pengujian kolinieritas dengan menghitung nilai Variance Inflation Factor (VIF). Nilai VIF digunakan untuk melihat apakah terjadi multikolinieritas atau tidak di dalam model sehingga nilai VIF harus lebih kecil dari 10 agar tidak ada indikasi terjadi multikolinieritas (Petrarca, Russolillo, & Trinchera, 2017). Pada hasil yang telah dilakukan, didapatkan hasil bahwa nilai VIF pada setiap variabel dalam model adalah 3,877 (SCA), dan 1,993 (BP).…”