Impulse Response in Eviews

[Seeking_Knowledge]

Hello,

what does the numbers on the Y-axis or Left axis mean?

[NicolasR]

Hi,

When you give an impulse to a system of equations or an equation, you do it through the residuals (also called innovations) which are the non explained part of the depent variable, so they are in units of the impulse variable. Usually the constant of the model is not taken into account in the impulse response exercise, because is interesting to see how the variable reacts instead of see how the level of the variable reacts.

[Seeking_Knowledge]

hi NicolasR thank you for your reply,

I'm not sure I have understood everthing, but it is a bit better,

Impulse Response in Graph mode

the horisontal line - x tells us about the number of ' days , months etc a shock is affecing the dependent variable.

1) what about the vertical line, y what are those values?

2) and what is the percentual value of a Cholesky SD adjusted

[NicolasR]

1) In units of the response variable.

2) Cholesky is a matrix decomposition that ensures that your residual covariance matrix could be transformed into a diagonal matrix. The purpose of this Cholesky decomposition is that the impulse to one variable (or their innovations) must be unrelated to the impulse in another variable, otherwise, it is unrealistic to assume that one variable would remain static (no impulse) while the other moves (impulse).

Best regards.

[Seeking_Knowledge]

thank you NicolasR,

1) units do I need to transform them to percentage in eviews, or can I interpret them as percentage values

2) I still don't understand what exactly the value of one Cholesky SD is, basically a standard deviation of what may I ask?

3) I want to induce a 1 % shock, should I choose Residual one unit?

[EViews Glenn]

http://faculty.chicagobooth.edu/jeffrey ... notes3.pdf

[Seeking_Knowledge]

That didn't really answer my questions......

I'm not sure whether you (Glenn) is tired of all these noob questions or maybe don't get paid for doing this..... :D

I need more practical answers, not something I could most likely google up myself.

It might be I don't understand it conceptually, but I do believe these are very specific questions.
So please can someone answer?

(A) I induce a one SD Cholesky shock in " sugar price variable " and the response value in units of " Cola share price variable " is 0.0017" at " day 5 - keep in mind all variables are in log form.
What is the value of SD change of sugar price in money? what does one SD change mean in the sugar price variable... should I take the square root of the sugar price ??

(B) Should I find look up the residual in the covariance matrix.... example: residual of sugar price variable = 0.00027 and divide 0.0017 / 0.00027 = 6.29 is this a 6.29 % in " Cola share price variable " ?
Is this correct methodology to find percentage change or should I just change native units to percentage in EVIEWS ??

[EViews Glenn]

Your questions are more econometric in nature and not EViews specific and are therefore a bit beyond normal technical support.

That said, to answer the EViews-centric question, the responses in EViews, and anything that refers to units is in the units of the original variable. Any economic interpretation beyond that is up to the user as we don't know whether your variables are in levels, logs, differences, or some other transformation. My linking to the PDF was in response to your last question about not understanding how to interpret the Cholesky option.

[Seeking_Knowledge]

Hello Glenn

1) I did state in last post, all variables are log forms, I will consider level if it makes it easier.

I changed the IRF from units to percent ( in log form ) and a typical response value changed from 0.02 ---> 0.2 %

Is this correct use of percentage ?

2) Also a Cholesky SD .....a SD of what ?? the mean price of a given shock variable or the residuals??????

I need to be able to explain this, I can't just say I induce a one standard deviation of shock from price variable X to response variable Y.
People will wonder and and ask standard deviation of what

These are very very Eviews specific questions, please helpe me

[startz]

If you are choosing "residual one unit" and all your variables are in logs, you are showing the of response log(y) to a one unit change in the residual. Since a unit change in the log is (roughly) a percentage change, you are showing the percentage response of a variable to a percentage change in the (not logged) residual.

[Seeking_Knowledge]

Hi

how much of a percentage change is a ONE SD CHolesky DOF ?

[trubador]

All of these questions are purely econometric in nature and are not directly related to EViews. So you are mistaken to think otherwise. You are simply having difficulty in understanding the mechanics of impulse response analysis. Although all your questions so far have been properly answered by NicolasR, Glenn and Startz, I will try to consolidate them as a final attempt:

1) Standard Deviation (SD) shocks are preferred, since they help align the responses in the proper order of magnitude (e.g. interest rates and gdp are measured in very different units).
2) Cholesky is a type of factorization that imposes a specific (triangular) form of structural model (i.e. first variable being the most exogenous).
3) A Cholesky SD is obtained through factorization of variance-covariance matrix of residuals, so the scales match to that of related dependent variables.
4) The responses are in the units of the dependent variables.
5) Impulse responses are linear in the sizes of shocks, so you can rescale them as you wish.
6) When you rescale the shock/impulse, you should do the same for corresponding response.

[usereviews]

Hi,
I have estimated a VAR and want to compute impuse responses with the Cholesky option. This option produces responses to a one standard deviation shock. I am interested in a one-unit shock rather than a one SD. How can I do the conversion? Where can I find the standard deviation of structural shocks? Thank you.

[mailin]

Hi,

Good day.

I would to ask on how to create permanent shocks in EVIEWS 8. I was looking at the guide under user specified impulse response function, but i think it only creates one time shocks. How can you make a permanent shocks let say 10% of the variable of interest.

Thanks.

Kind regards,

Mai Lin