this update fixes the simulator model used in stochastic volatility example and introduces some updates in the inference task created with get_model.
about the simulator:
update adds exp and square root transformation so that log-returns are calculated as np.exp(0.5 * x_t) * v_t as proposed in previous works, and also updates log-volatilities initialisation. the previous version sampled x(0) from ss.norm.rvs(mu + phi * (- mu), sigma), which corresponds to assuming that x(-1) = 0. this initialisation option is still available in the current version, ie the user can provide a previous x value and can choose to provide zero, but the default initialisation now uses the equilibrium distribution ss.norm.rvs(mu, sigma/np.sqrt(1 - phi**2)) as proposed in vankov et al (2019).
about the task:
the inference task created in get_model is based on the example in priddle and drovandi (2020) where the unknown parameters include the stable distribution parameters alpha and beta, and other parameters are assumed known and fixed. the fixed parameters are now set in get_model so that we can use alpha_stochastic_volatility_model to simulate data with other parameters also. in addition the parameter values mu=5 and phi=1 are substituted with mu=0 and phi=0.95. i chose mu=0 because mu is not used when phi=1 and phi=0.95 because it is close to 1 but complies with the model constraint -1 < phi < 1. also since alpha and beta control tail heaviness and skewness, i added quantile-based kurtosis and skewness as summaries for distance-based methods.
Please make sure
[x] You have updated the CHANGELOG.rst
[x] You have provided a short summary of your changes (see previous section)
[x] You have listed the copyright holder for the work you are submitting (see next section)
If your contribution adds, removes or somehow changes the functional behavior of the package, please check that
[ ] You have included or updated all the relevant documentation
[ ] You have added appropriate unit tests to ensure the new features behave as expected
and the proposed changes pass all unit tests (check step 6 of CONTRIBUTING.rst for details)
Copyright and Licensing
Please list the copyright holder for the work you are submitting (this will be you or your assignee, such as a university or company):
By submitting this pull request, the copyright holder is agreeing to license the submitted work under the following licenses:
Summary:
this update fixes the simulator model used in stochastic volatility example and introduces some updates in the inference task created with
get_model.about the simulator:
update adds exp and square root transformation so that log-returns are calculated as
np.exp(0.5 * x_t) * v_tas proposed in previous works, and also updates log-volatilities initialisation. the previous version sampled x(0) fromss.norm.rvs(mu + phi * (- mu), sigma), which corresponds to assuming that x(-1) = 0. this initialisation option is still available in the current version, ie the user can provide a previous x value and can choose to provide zero, but the default initialisation now uses the equilibrium distributionss.norm.rvs(mu, sigma/np.sqrt(1 - phi**2))as proposed in vankov et al (2019).about the task:
the inference task created in
get_modelis based on the example in priddle and drovandi (2020) where the unknown parameters include the stable distribution parameters alpha and beta, and other parameters are assumed known and fixed. the fixed parameters are now set inget_modelso that we can usealpha_stochastic_volatility_modelto simulate data with other parameters also. in addition the parameter valuesmu=5andphi=1are substituted withmu=0andphi=0.95. i chosemu=0becausemuis not used whenphi=1andphi=0.95because it is close to 1 but complies with the model constraint-1 < phi < 1. also since alpha and beta control tail heaviness and skewness, i added quantile-based kurtosis and skewness as summaries for distance-based methods.Please make sure
If your contribution adds, removes or somehow changes the functional behavior of the package, please check that
and the proposed changes pass all unit tests (check step 6 of CONTRIBUTING.rst for details)
Copyright and Licensing
Please list the copyright holder for the work you are submitting (this will be you or your assignee, such as a university or company):
By submitting this pull request, the copyright holder is agreeing to license the submitted work under the following licenses: