# SVAR A/B Models

I am currently working on SVAR models in R to add a theoretical background to my VAR model. I want to use the vars package for this. I came across a statistical question. With the command SVAR 3 different models can be calculated: The A-model, the B-model and the AB-model. I understand how the methods differ technically, but I find it difficult to interpret the methods.

$A {y}_t = A_1^*{y}_{t-1} + … + A_p^*{y}_{t-p} + B{\varepsilon}_t$

According to Pfaff 2008, the A model has restrictions in the dependent variable and the B model sets the restrictions in the residuals. Accordingly, I suspect that the B-model can be controlled like an exogenous shock to the model. (In the first period.) The restrictions in A control how a shock affects the further course? Am I right about that? Is there a practical tip as to when which model should be used?