Hello all,
I am attempting to computer IRF graphs for the variables that comprise factors in a 2-step factor-augmented vector autogression. The IRF GRAPH function in Stata will not allow me to do this - it will only create IRF graphs for the factors (responding to some policy innovation) rather than for the variables comprising my factors.
I think the solution here is to use the factor loading for each variable from each factor and also use the irf tables for each factor to create a linear combination of values that show me how each variable responds to a policy innovation and then graph these values over time. This should produce IRF graphs.
So, I have two sets of tables. The first table contains factor loadings for each of my variables (I have about 90 variables and 3 factors - so 3 factor loadings for each of those 90 variables).
Here is an example of my table containing factor loadings:

Next, I have a table that contains the magnitude of how each of my three factors responds to a policy innovation over time:
I use this code: irf table sirf, set(var1.irf) impulse(FEDFUNDS) response(FI1 FI2 FI3) std
To produce this table:

Now, what I need to do is somehow store the factor loadings for each variable so that I can multiply those factor loadings for each variable with their respective responses to innovations in the IRF table. Repeating this for all of my variables and graphing should produce my desired graphs.
An example:
Consider variable E. If I want to graph how my variable E responds to an innovation in FEDFUNDS I should have:
E_t1 = (-.002676 *.9639) + (.012273 *.0073) + (.0046*.016512)
E_t2 =...
.
.
.
E_t10....
Does anyone have a good idea about how I can store the values from these tables into vectors that would make such an exercise less painless than just a brute force/manual approach?
I am attempting to computer IRF graphs for the variables that comprise factors in a 2-step factor-augmented vector autogression. The IRF GRAPH function in Stata will not allow me to do this - it will only create IRF graphs for the factors (responding to some policy innovation) rather than for the variables comprising my factors.
I think the solution here is to use the factor loading for each variable from each factor and also use the irf tables for each factor to create a linear combination of values that show me how each variable responds to a policy innovation and then graph these values over time. This should produce IRF graphs.
So, I have two sets of tables. The first table contains factor loadings for each of my variables (I have about 90 variables and 3 factors - so 3 factor loadings for each of those 90 variables).
Here is an example of my table containing factor loadings:
Next, I have a table that contains the magnitude of how each of my three factors responds to a policy innovation over time:
I use this code: irf table sirf, set(var1.irf) impulse(FEDFUNDS) response(FI1 FI2 FI3) std
To produce this table:
Now, what I need to do is somehow store the factor loadings for each variable so that I can multiply those factor loadings for each variable with their respective responses to innovations in the IRF table. Repeating this for all of my variables and graphing should produce my desired graphs.
An example:
Consider variable E. If I want to graph how my variable E responds to an innovation in FEDFUNDS I should have:
E_t1 = (-.002676 *.9639) + (.012273 *.0073) + (.0046*.016512)
E_t2 =...
.
.
.
E_t10....
Does anyone have a good idea about how I can store the values from these tables into vectors that would make such an exercise less painless than just a brute force/manual approach?
Comment