MA Support, Intercept Flexibility, and Seasonal Interpretation #4
Description
Dear Developers,
I hope this message finds you well. I have been using your R package for Markov Switching Autoregressive modeling and greatly appreciate the functionality it provides. I would like to offer some suggestions that I believe could significantly enhance its flexibility and usability for time series analysis.
Support for MA Terms / MS-ARIMA Models
Currently, the package supports autoregressive (AR) components but does not allow for moving average (MA) terms. In practice, I often encounter time series with residual autocorrelation and heteroskedasticity, even after increasing the AR lag. Unfortunately, since MA terms are not supported, I am unable to address these issues effectively. This also limits the integration with models like GARCH, which could otherwise help address heteroskedasticity if MA were included. Adding MA support or enabling MS-ARIMA structures would be extremely valuable.
Option to Remove Intercepts
I would also like to request an option to remove the regime-specific intercepts. In models with seasonal or other dummy variables, the intercept tends to absorb much of the variation, making it difficult to observe significant regime-specific changes in the coefficients of interest. Removing the intercept would make it easier to interpret parameter changes across regimes—especially in models focused on monthly or seasonal dynamics.
These enhancements would make the package more versatile and allow for richer interpretation, especially in applied economic or environmental time series modeling.
Thank you for your time and consideration. I would be happy to discuss this further or contribute to testing if these features are considered for future development.
Best regards,
Sandy