search

ssmit1986 / BayesianInference

Wolfram Language application for Bayesian inference and Gaussian process regression
MIT License
37 stars 4 forks source link

Fitting with time series processes #4

Closed YetAnotherDima closed 4 years ago

YetAnotherDima commented 4 years ago

In "example_code.nb" for "Fitting with time series processes" you use this block: With[{mu = 0, logsigma = -4}, Print@obj["LogLikelihoodFunction"][{mu, logsigma}]; Print@LogLikelihood[ GeometricBrownianMotionProcess[mu, Exp[logsigma], start], ts] ]

Why you use logsigma = -4?

ssmit1986 commented 4 years ago

Hi, Dima,

That block of code is just for the purpose of illustrating that the likelihood function defined in that example isn't the same as the one computed by using: 

LogLikelihood[GeometricBrownianMotionProcess[...], ...]

That particular section is just an introduction to time series modelling. The model defined by using "GeneratingDistribution" -> process[t] assumes that all observations in the time series are independent, which is obviously not very realistic. The value of logsigma = -4 is just to illustrate that the two log-likelihoods are different; nothing else. It could've been any number, really.

YetAnotherDima commented 4 years ago

Thank you