I’m trying to implement a multivariate stochastic volatility model in PyMC3. The model consists of two correlated SDEs: the return process (observed) and the volatility process (unobserved). I basically need to extend either the EulerMaruyama timeseries class to handle a bivariate SDE, or, the MvGau…

I think we had it right the first time. error_s is a just a vector of i.i.d standard normals error_s = pm.Normal('error_s',mu=0.0,sd=1, shape=len(returns)) Though I’m still getting an error about input dimension ValueError: Input dimension mis-match. (input[0].shape[0] = 3812, input[1].shape[0] =…

I got it to run but the results don’t make sense. I don’t think this is the right way to solve the problem. The additional normal random required to pass in the observed returns is undesirable. There’s got to be a way to write a multivariate class to support this import numpy as np import matplotli…

[image] simon_o: Yes, unfortunately I’m not sure why PyMC3 won’t allow that, and I’m not aware of a workaround (without introducing superfluous noise). That’s why I added the third Brownian motion. Adding additional noise is the workaround, so your solution makes sense to me. [image] Mari…

[image] simon_o: Yes, unfortunately I’m not sure why PyMC3 won’t allow that, and I’m not aware of a workaround (without introducing superfluous noise). That’s why I added the third Brownian motion. Perhaps giving it a really low standard deviation is enough? But I’ll let others who are more know…

I may have found a way to workaround the problem but am getting errors. I defined \epsilon_t^s b using pm.Normal like before. I pass this into the drift coefficient of the variance process SDE, as before. For the return process, I can’t use pm.Deterministic because I can’t pass in the observed retur…

[image] Marion2025: Is there a reason why pm.Deterministic does not allow you to pass in an observed variable? Are there any workarounds? pymc3 needs a likelihood function to sample, and if you observed a function of some random variables (ie., observing a Deterministic) the likelihood expres…

I tried implementing the model without correlation as a starting point but am getting a “Bad Initial Energy” error. I tried setting init=‘adapt_diag’ in pm.sample but still get the error. I even tried changing my priors to ensure they are positive. import numpy as np import matplotlib.pyplot as p…

Ok I think I solved that issue…need to ensure Vt is positive when taking the square root. Now I just need to get the correlation worked in import numpy as np import matplotlib.pyplot as plt import pymc3 as pm from pymc3.distributions.timeseries import EulerMaruyama returns = np.genfromtxt(pm.get_d…

Even with that fix, the output doesn’t make any sense. [image]

Help with Multivariate SDE Timeseries

Questions

Marion2025 September 15, 2018, 9:08pm 23

Any help on this? I’m at a dead end.

Product of two probabilities distributions

Topic		Replies	Views
Stochastic Volatility with Jump Diffusion? Questions	24	5283	April 2, 2024
Error in Euler Maruyama Questions	0	319	September 27, 2021
Stochastic volatility jumps Questions	5	1146	January 6, 2022
Estimate parameters of two correlated processes Questions	4	1298	February 10, 2020
Euler Maruyama issue v5 modeling	11	698	October 6, 2023

Help with Multivariate SDE Timeseries

Related topics