0
$\begingroup$

I'm using MSGARCH package in R. By return_data/Volatility(fit.model), I get the residuals. When I calculate the standard deviation of the residuals, it turns out that it's close to 1 for all residuals. However, the standard deviation in each regime differs greatly. It's about 0.7 in one and 1.6 in the other.

Is this reasonable? I imagine that the residual would have unit variance in each regime respectively, since in each regime it's filtered by a GARCH model.

This result can be reproduced by any data I tried, so it's more about how the function Volatility works.

$\endgroup$
5
  • $\begingroup$ Hi: s the garch model the same in each regime ? what differentiates each regime ? $\endgroup$ Commented Aug 15, 2020 at 14:30
  • $\begingroup$ The parameters differ in each regime. The switch is governed by a Markov chain specified in paper A New Approach to Markov-Switching GARCH Models. $\endgroup$ Commented Aug 15, 2020 at 14:38
  • $\begingroup$ I don't understand what you mean when you say that it's 1.0 for all residuals but 0.7 in one regime and 1.6 in the other. Do you mean it's 1.0 when you don't consider regime and just calculate it over the whole series ? Also, are these calculations on the residuals done after the garch fit ? $\endgroup$ Commented Aug 15, 2020 at 21:43
  • $\begingroup$ Yes. The calculations here are all on the residuals. The standard deviation of all residuals is close to 1. But when I separate the residual series according to regimes, the standard deviation differs a lot, which I thought would be still close to 1 respectively. $\endgroup$ Commented Aug 15, 2020 at 23:06
  • $\begingroup$ Hi: I've never worked with regime switching but, since you've seperated things into two regimes, I wouldn't worry about what the sd is when you don't consider regimes. It doesn't really have any purpose in your framework. $\endgroup$ Commented Aug 16, 2020 at 19:07

0

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.