EViews.com

That option is not yet available. I will try to do it.

Hi dakila,
Thank you so much for your add-in!
Is it possible to add more than one observable variable in the FAVAR add-in?
For example: There are two observable variables- Y1 and Y2.
I try to obtain ① IRFs between Y1 and Y2, ②IRFs of some variables in X in response to Y1, ③IRFs of some variables in X in response to Y2,
④variance decomposition of some variables in X considering Y1 and Y2 separately.
I hope my description doesn't confuse you.

Could you give me some advice on how to obtain these IRFs and variance decomposition? Thank you very much.

No problem, you can add more than one observable variable. For example,

favar(factor=3,horizon=48,rep=1000,ci=0.9,save=1,vd=1) 13 xdata xslow xir tcode yx_name @ ip cpi ffr

dakila wrote:No problem, you can add more than one observable variable. For example,
favar(factor=3,horizon=48,rep=1000,ci=0.9,save=1,vd=1) 13 xdata xslow xir tcode yx_name @ ip cpi ffr

Hi dakila,
Thank you for your response!
I tried to include 2 observable variables. But I only got an IRF figure and a variance decomposition table.
I have no idea which observable variables cause the shock. Could you help me to explain the results? Are there any mistakes in my command?
Thank you very much.


command
favar(factor=3,horizon=48,rep=100,ci=0.9,save=1,vd=1) 13 x xslow xir tcode yx_name @ series77 series16

No mistakes. Sorry, It will show you IRF figure and variance decomposition of the last variable (shock to series 16). I will try to include IRF figure of other variables.

Thank you very much for your help. I look forward to the update！

Hi dakila,
If I set "xslow" equal to "xdata" in the FAVAR add-in, does it mean that I extract principal components from the whole information set "xdata" without distinguish fast-moving variables and slow-moving variables?
(xdata:group of X variables; xslow:group of slow-moving variables)

If so, is it uncompatible with the recursive ordering? (order observable variable last)
Could you help me to figure out these questions? Thank you

Hi Briskair,

The favar add-in is updated. Now it includes the optional impulse variable. For example,

Hi;

Thanks for the very useful add-in. I want to ask related to impulses responses. I try to analyze oil effects on the economy. But, responses of the factor series to oil vary according to the method which is used. Generalized IR has significantly different results. Which of them do you suggest to use?

Finally, is there possible to obtain the generalized responses of special variables?

Thanks..

dakila wrote:Hi Briskair,

The favar add-in is updated. Now it includes the optional impulse variable. For example,

Code: Select all

favar(factor=3,horizon=48,rep=1000,ci=0.9,impulse=ffr) 13 xdata xslow xir tcode yx_name @ ffr


Thank you so much for your update!

Hi dakila,
How could I use information criteria of Bai&Ng（2002) to determine the number of factors?
I would greatly appreciate if you could help me!

Hi dakila;

Thank you for your all kindly help. Finally, I want to ask that is it possible to add exogenous variable in the FAVAR model and derive the responses of some observables to change in the exogenous variable?

I work with oil price as a dependent variable. And I want to restrict feedback from endogenous to oil.

dakila wrote:No. All interpretations are in std. its does not matter whether the horizon is 48 or 60. The result is the same.

Thank you so much for sharing the FAVAR add-in. Given the replication file for Bernanke et al. (2005), it is by far the most intuitive package for FAVAR analysis I have encountered so far.

I would like to use the add-in to compute impulse responses and variance decompositions for three observable policy variables.

For the variance decomposition I would like to compare the contribution of the policy variables under consideration at different time horizons (6 months, 12, months, ..., 60 months). However, from your post it seems that the variance decomposition will always yield the same results no matter the chosen time horizon. Is there a way to disentangle the contributions of the different shocks over time?

Thanks in advance. Your help is much appreciated.

Generally It will depend on the time horizons. That case the variance decomposition converged quickly the long run value.

dakila wrote:Generally It will depend on the time horizons. That case the variance decomposition converged quickly the long run value.

Dear dakila,

thanks so much for your timely reply. So in other words, given a different data set, we may very well observe a difference between the fraction of the variance of the forecast error explained by the monetary policy shock at different time horizons.

Given your example code from Bernanke et al. (2005), I suppose that we should use the commands


and


to obtain and compare the variance decomposition with respect to a monetary policy shock over six and 60 months, respectively. Is that right?

If time permits, I would also like to know if there is a command to scale the impulse response functions rather than use the drop-down menu.

Thanks in advance!