[Gretl-users] Re: Some new features in the next version

Am 06.12.2021 um 09:23 schrieb Olasehinde Timmy: This remains unlikely; the main reason (at least from my point of view) being the following argument from Kilian & Lütkepohl (2017, SVAR Analysis, chapter 18, pp. 617-618), with which I agree: "... Pesaran and Shin (1998) explicitly advocate the use of the generalized impulse responses ... as being more general than traditional structural VAR impulse responses... They suggest that studying the response of the model variables to reduced-form shocks is more general because it avoids the use of economic identifying assumptions that may be controversial. This argument is not persuasive because reduced-form shocks have no economic interpretation and violate the ceteris paribus assumption required for quantifying causal relationships in the data. At best, these responses may serve as a description of the properties of the data. They are not informative about the structural model. Moreover, if we were actually interested in responses to reduced-form shocks in linear VAR models, such responses could easily have been generated using the standard tools..." In gretl's context, these standard tools are for example the vma() and irf() functions which allow you to calculate the reduced-form impulse responses. However, it does seem to be true that there is no direct option to get the reduced-form IRFs from the GUI. (Because typically they are not of direct interest.) If you really must see the reduced-form IRF related to the shock variable x, here's what you can do in the GUI as a workaround: Use the Choleski scheme and move x to the extreme position such that its shock only affects its own equation on impact, nothing else. cheers Sven