Closed hannesdatta closed 4 years ago
Hi Hannes, thanks for all your help. We've now added a noLDV option in the newest version of dynamac (v 0.1.11), which should be up on Github very soon, and CRAN a few days after that. Thanks!
Hey everybody, thanks again for incorporating my request.
One question though. The warning is quite explicit, and the option is even not recommended.
Why issue such a drastic warning, if all this flag achieves is to estimate (the very legit) formula (11) in Philips 2018, i.e., a model WITHOUT lagged dependent variable, but a DV in first-differences?
This model is suggested by Andrew in the case of no cointegration (Figure 1, p. 233).
What I need is a clarification from anybody (Andrew? Soren?), confirming that using the new option noLDV
is indeed in line with case (r) / formula (11) in Philips 2018.
If this is indeed true, then let us rather reference the paper in the documentation, as opposed to issuing this big warning which may prevent users from estimating the model that was suggested by Andrew in the first place ;)
REFERENCES: Philips, A. Q. (2018). Have your cake and eat it too? Cointegration and dynamic inference from autoregressive distributed lag models. American Journal of Political Science, 62(1), 230-244.
Dear Soren and Andrew,
I am interested in estimating case (r) from Andrew’s 2018 paper (“first difference DV, and run ARDL(p,q)”). Specifying EC = TRUE is not enough, as dynardl issues a warning and automatically adds the lagged dependent variable to the model. If my reading is correct, dynardl covers case (j) (estimate in levels), and case (o) (conclude cointegration). Case (r) is not covered.
As the current implementation, to the best of my knowledge, does not allow the user to estimate formula (11) (or, in other words, formula 8 – the EC model – with theta_0 set to zero), I have experimented with adding an option to dynardl that does allow the user to do just that.
I’m adding a PR to your repository with my (very experimental) updated code. Also find attached markdown which illustrates the new functionality.
Once again, thanks for your fantastic work!