High Dimensional Model Representation (HDMR) is becoming a powerful tool for multivariate analysis in recent years as researches go on. It is basically an expansion in ascending multivariance and its most important aspect is the possibility of truncating it at univariate level. This truncation approximation works well as long as the univariate behaviour of the original function dominates. If it does not dominate then one need to seek certain ways to increase the dominancy of the univariance. One of the recently noticed way, is to use not the original function in HDMR but its image under an appropriately chosen transformation. The choise can be realized in such a way that the resulting HDMR's univariance dominates. This work focuses on these issues at phenomenological level although certain instruction will be given for numerical implementation in the presentation.