Congratulations on a great website so far! I enjoyed going through your work (particularly your neatly coded functions of which I am a fan) and learned a lot about Bayesian forecasting. I am grateful that Tomasz assigned me to be your peer reviewer. :)
My suggestions are incorporated in the commit 6806f7acba5a66853f6d09fc15f9243e0001b191. Below is a summary:
[x] I incorported some suggested revisions to the text. These are mainly cosmetic and do not change the story of the paper.
[x] Inclusion of an interactive 3D plot of the forecast densities. I included this since one of the comments during the presentation was that the general direction of the forecast was not as apparent from the provided vantage point of the current plot. The 3D surface plots generated by plot_ly are interactive and can be moved around by users (and they look kind of cool), which allows for viewing the densities in different angles. You can even view the 3D plot from the bottom to get a 2D plot that clearly shows the direction of the forecast across time. I implemented these for the basic model forecasts for now (using colors from our subjects' palette), but it should be easy to adopt to the extended model as well should you see them fit to be incorporated them in your website.
[x] General comment: Since one of the questions the paper aims to address is "if/when will inflation in Australia return to the 2-3 percent target", I was wondering if it would be prudent to just forecast inflation directly as opposed to CPI (which is a stock value that generally tends to go up as long as inflation is postive, even if at the target rate). I think this could be implemented in a relatively straightforward manner, just by applying an fd(log(cpi)) transformation to your current CPI variable (you'll lose one observation though so removing one as well from the other variables will be necessary).
Please reach out to me any time if you have questions or clarifications. Thank you again!
Hello @mandyxmg
Congratulations on a great website so far! I enjoyed going through your work (particularly your neatly coded functions of which I am a fan) and learned a lot about Bayesian forecasting. I am grateful that Tomasz assigned me to be your peer reviewer. :)
My suggestions are incorporated in the commit 6806f7acba5a66853f6d09fc15f9243e0001b191. Below is a summary:
plot_lyare interactive and can be moved around by users (and they look kind of cool), which allows for viewing the densities in different angles. You can even view the 3D plot from the bottom to get a 2D plot that clearly shows the direction of the forecast across time. I implemented these for the basic model forecasts for now (using colors from our subjects' palette), but it should be easy to adopt to the extended model as well should you see them fit to be incorporated them in your website.fd(log(cpi))transformation to your current CPI variable (you'll lose one observation though so removing one as well from the other variables will be necessary).Please reach out to me any time if you have questions or clarifications. Thank you again!
Best regards, Ray