search

alfsn / regime-switching-hmm

0 stars 0 forks source link

svar_type #20

Closed alfsn closed 9 months ago

alfsn commented 10 months ago

The svar_type argument in a Structural Vector Autoregression (SVAR) model specifies the methodology used to estimate the structural shocks and identify the causal relationships between variables. Different SVAR types make different assumptions and employ different identification schemes. Let's explore the differences in commonly used svar_type options:

  1. "AB" (Andrews and Bruce): The "AB" SVAR type assumes that the errors are contemporaneously uncorrelated but may have lagged relationships. This means that in this SVAR type, contemporaneous (same-time) shocks are assumed to be uncorrelated between variables, but lagged (past-time) relationships are allowed.

  2. "LR" (Lütkepohl and Reimers): The "LR" SVAR type is based on the work of Lütkepohl and Reimers. It allows contemporaneous and lagged relationships among the variables, and it estimates both the contemporaneous and lagged structural shocks.

  3. "FA" (Frisch and Blanchard): The "FA" SVAR type, based on the work of Frisch and Blanchard, assumes that only the current values of the variables are impacted by the structural shocks, and past values are not affected by the shocks. This implies that there are no lagged relationships.

  4. "H" (Hirano and Porter): The "H" SVAR type, developed by Hirano and Porter, allows for both contemporaneous and lagged effects of structural shocks. This means it allows for shocks to contemporaneously impact variables and to have lagged effects as well.

  5. "R" (Ricardian): The "R" SVAR type is based on the work of Blanchard and Perotti and assumes that current and future values of the variables are impacted by structural shocks, but past values are not affected by the shocks. This implies that there are no lagged relationships.

The choice of SVAR type is essential because it determines the structural assumptions you make about the relationships between variables and the nature of the structural shocks. Different SVAR types may be more appropriate for different economic or research scenarios. The choice may also depend on the specific assumptions and theories you want to test or the characteristics of your data. It's important to carefully consider the underlying assumptions of the SVAR type and choose the one that best matches the economic or empirical context of your study.

alfsn commented 10 months ago

Por ahora voy con AB

alfsn commented 10 months ago

De baja porque para la prediccion solo necesito la forma reducida del VAR, y esto implica que las restricciones estructurales no son relevantes para predecir.