I needed to define a subset of autoregressive lags (set some of them to zero) on both your ARMA and VAR implementations for a project I was working on, so I defined a parameter that specifies those. Since it is done, I thought it would be good to incorporate that adjustment on your package. Beyond that, I tweaked the code a bit for it to be easier to follow. One important adjustment I did though is to initialize the negative log-likelihood, which was the reason why I was getting some funky variations in my results across runs. Thanks for sharing your code!
Hey Eduardo - thanks for the pull request! I haven't had a chance to look at the library for a while, so I will look this weekend, including your PR. Thanks for the contribution!
Hi Ross,
I needed to define a subset of autoregressive lags (set some of them to zero) on both your ARMA and VAR implementations for a project I was working on, so I defined a parameter that specifies those. Since it is done, I thought it would be good to incorporate that adjustment on your package. Beyond that, I tweaked the code a bit for it to be easier to follow. One important adjustment I did though is to initialize the negative log-likelihood, which was the reason why I was getting some funky variations in my results across runs. Thanks for sharing your code!