No, PLS-SEM does not assume linearity nor data normality.

Partial least squares structural equation modeling (PLS-SEM) does not require these assumptions to be satisfied, as it is a nonparametric method that can handle complex and nonlinear relationships among latent variables. However, some researchers suggest that checking the normality and linearity of the data can help to assess the quality and validity of the model, especially when comparing it with other methods such as covariance-based SEM. Therefore, it is advisable to perform some exploratory data analysis and visual inspection of the data before running PLS-SEM, especially when dealing with a large and heterogeneous sample of 600 observations.PLS-SEM does not require these assumptions to be satisfied, as it is a nonparametric method that can handle complex and nonlinear relationships among latent variables. However, some researchers suggest that checking the normality and linearity of the data can help to assess the quality and validity of the model, especially when comparing it with other methods such as covariance-based SEM. Therefore, it is advisable to perform some exploratory data analysis and visual inspection of the data before running PLS-SEM, especially when dealing with a large and heterogeneous sample of 600 observations.