I performed a 2D parametric analysis where one variable is much larger than the other. Basically I sampled a function in many points: let's say 5 points for $x_1$ and 5 points for $x_2$, where the domain of $x_1$ is wider than the one of $x_2$.
When I try to obtain the weights, the interpolation does not make sense at all.
My guess is that since the function is radial, it acts in the same way along $x_1$ and $x_2$, but this could not approximate the underlying function well because I expect a similar variation of the function in the sampled domain in any direction.
How is this problem treated with RBF interpolation?
My first approach would be to scale both variables to "normalize" them, so that the RBF would behave in the same way in both directions, but I am not sure that this would be the best way to deal with this.