Functional Data Analysis (FDA) has recently made considerable progress because of easier access to the data that are essentially in the form of curves. Modeling of Iranian precipitation based on temperature and humidity with continuous the essential nature of such phenomena that are continuous functions of time has not been done properly. The corresponding data are generally collected daily or monthly (discretely). However, if one treats those data as multivariate observations and analyzes by multivariate methods, it may cause some problems such as infinite number of solutions for normal equations in regression.

In regard to the fact that the original discrete data must be firstly converted to continuous functions, we usually use “basis function methods” for dimension reduction of the data due to its simplicity. In this article, problems arising form using multiple regression methods instead of applying functional regression approaches are discussed.

We have treated a real dataset that was collected from 102 Iranian weather stations in 2006. The dataset ... [To continue please click here]