# Why do Gaussian copula does not have a closed form? hence, why numerical estimation is needed?

I am working on Gaussian copula. I always read that, Elliptical copulas do not have closed form expression and hence, the numerical estimation is needed. I really do not understand what does closed form means for Gaussian copula. And Why it does not have? Also, what is the relationship between closed form and numerical estimation?

Any help, please?

• That copula function has inverse normal CDFs as arguments of a multivariate normal CDF, so this could be what you're looking for: stats.stackexchange.com/questions/18286/… Commented Oct 1, 2018 at 7:33

## 1 Answer

A closed form means that there is a defined number of solutions that solve the problem. https://en.wikipedia.org/wiki/Closed-form_expression

The issue with two Gaussian variables is that they give a circular joint probability for which rotation is arbitrary, so there is no single solution.

An ellipse can be considered simply a circle where one axis is scaled differently, but this scaling can be applied to any arbitrary rotation of the unit vector relationship between them.

Basically you are looking at the flip side of the issue covered in several questions on Independent Component Analysis.

What is mean by the non-gaussianity in the independent component analysis(ICA)?

The Independence in Independent Component Analysis - Intuitive Explanation

Numerical estimation is used to provide a way to break the impasse of an open solution. We can't identify finite (or manageable) unique solutions analytically therefore we apply some decision rubrics to allow us to move forward. Its a compromise that accepts we would waste eternity trying to find the true solution, but we need something we can act on in a finite period of time.

What those rules look like depends on what matters to the analyst and likely some will be completely arbitrary. The key is that we accept a certain risk that our chosen solution is sub-optimal to allow us not to become frozen in indecision.

• Thank you for your answer. However, then why the numerical estimation is needed? Commented Oct 1, 2018 at 11:29