Introduction to Copulas

[@tiankhean]

Great video. For the first time in my life, I truly understand Copulas. I am 74 years old.

[@adamfattal9602]

Hopefully I'll truly understand them too by the time I'm 74 years old, although it's not looking likely.

[@itooflemma]

@@adamfattal9602lmao

[@MJtheFellowActuary]

This video is just one of many in a paid Udemy Course. To see the rest, visit this link: www.udemy.com/course/introduction-to-copulas/?referralCode=FD09B59500CFA149839D

[@harshagarwal4866]

very easily explained. I have made a model for my company to calculate operational risk capital using copulas

[@MJtheFellowActuary]

That sounds cool!! If you ever share it on the internet, please send me a link

[@ankitapanwar7633]

Hello Harsh, can you please help me with the Gaussian copula.

[@sohamde3331]

Great video thanks. I used recently COPOD which is a copula based method for Outlier/Anomaly detection. and needed a background (like this video) on Copulas. Such a interesting and useful mathematical object!

[@avibank]

Very nice. Coming from a background in statistical methods in computer vision where multivariate/joint distributions are used extensively. I feel like this came up somewhere in the textbook but wasn't named.

[@lordpilot]

Hey, I'm trying to verify the inverse you highlight at 07:08, you write psi^(-1)(F(x)) = exp{ [-F(x)]^(1/alpha) }, however, I get psi^(-1)(F(x)) = exp{ - [F(x)]^(1/alpha) }. My procedure is by defining f(x) = x^alpha, g(x) = -x, h(x) = ln(x) such that psi(x) = (f o g o h)(x) Now using this formulation I can write the inverse as psi^(-1)(x) = (f o g o h)^-1(x) = (h^-1 o g^-1 o f^-1)(x) = exp{ - [x]^(1/alpha) }. Am I applying some rules wrongly? It seems by the used notation that you get psi^(-1)(x) = (f o g o h)^-1(x) = (h^-1 o f^-1 o g^-1)(x) = exp{ [-x]^(1/alpha) }

[@Achrononmaster]

You are in agreement. "MJ" used sloppy ^ notation and dropping your { } bracket. He's saying raise everything after the ^ to the _e_ power, which is evident in the fact he did not put it all in superscript position.

[@spooros]

great video, it would be very interesting if you make a video showing the calculation of SCR using copulas in R

[@shanikhan783]

This videos is very informative and actully give the information about the roots of Copulas. Thankuuu

[@aj_actuarial_ca]

Great video

[@Asd-jr8xj]

Thanks for explaining copula. Question: understood why the gumbel copula is important between CDR and market risk, but what about interdependencies between market risk and premium risk would you use gumbel copula to define correlation or normal one because one need to assume some kind of correlation but that might be not as strong correlation as it is in between cdr and market risk

[@giovannistephens3107]

Brilliant explanation! Thanks for sharing

[@Kaassap]

Great video. I know In late to the party but what book would you recommend for EVT a) rigour in underlying maths, and b) applications?

[@Achrononmaster]

@11:38 I agree it was a lot of greed. But any honest actuary was free to write about the more accurate Gumbel Copula. Not many did because they did not understand the fragility, and hence did not know the τ or α value. That in turn is likely because they mostly use neoclassical macro models. The instability of credit-driven capitalism is upwards. You don't know it was fragile (if you are a neoclassical poodle) until it's on the way down. They still haven't learned the lesson today. Only the smarter and more conservative actuaries are pricing rick correctly. How do you know derivatives like MBSs and CDOs are fragile without a macro model that understands the role of credit? You need to know the "mechanics" of the whole system in the macro (so basically MMT), not just past statistics.

[@Achrononmaster]

The instability was understood well before the short sellers got the bankers hooked. They were only looking at the rising defaults. A macro model seeing the credit rate of change "knew" the instability well before that short selling. Very, very few (count them on your hand) were doing such macro modelling. Most were DSGE whores or even worse SVAR kiddies, so basically frauds.

[@anormalID]

jeez this video is awesome

[@tomw4688]

Copula is the copulation between marginal distributions?

[@lucysteve5319]

Very nice video!

[@auntiewong6326]

so clear!!! Thank you