weird impulse response

[TomDoan]

Doesn't that give you an error message? That's the wrong syntax for print. You want

print / irf1(1,2)

[indrani_5]

How can I calculate something as basic as this in RATS?

IRF1(1,2)/0.000431707

Using the following command yields errors

set mm=IRF1(1,2)/0.000431707

## SX22. Expected Type SERIES[REAL], Got RECTANGULAR[REAL] Instead
>>>>1(1,2)/0.000431707<<<<

or should i just take out the data into excel or eviews and calculate BASIC statistics. That will be less time consuming I guess.

[TomDoan]

If you forget the syntax of a command, you can always look it up. You need a space before the = in SET.

set mm = IRF1(1,2)/0.000431707

[TomDoan]

That scaling factor looks rather small for a standard error.

If you're concerned about the sizes of the responses (again, what you're doing is not a good idea) use 100*log(x) rather than just log(x). That's standard practice among experienced VAR practitioners. (Note that that's what BQ do in the original paper).

[indrani_5]

I am following the method of Bullard and Keating (1995). They report standardized impulse responses whereby impulse response of each variable are divided by the standard error for that variable's residual. Is what I am doing the right way to get close to what they are doing?

[TomDoan]

That's a paper written probably 25 years ago, and in subsequent work with the same basic model, they did not do that:

Keating, John W & Nye, John V, 1998. "Permanent and Transitory Shocks in Real Output: Estimates from Nineteenth-Century and Postwar Economies," Journal of Money, Credit and Banking, Blackwell Publishing, vol. 30(2), pages 231-251, May.

It makes the calculation more complicated and the results murkier. (People do not think of GDP growth or inflation or unemployment rate in standard deviations). If you really want to do that (and ignore 25+ years of practice), then yes, you have to rescale the responses and upper and lower bounds. Just make sure you scale by a standard deviation and not a variance (as I said, that number doesn't look right for a standard deviation).

[ateeb]

How can we control the spacing between the values on vertical axis?

example: if the highest value is 1 and lowest is 0, the RATs automatically may plot with a space of 0.2 which means the y axis will have 0, 0.2, 0.4, 0.6, 0.8 and 1.0 - However if i want to plot it in a way that it plots with a gap of 0.1 like 0, 0.1, ... , 0.9 , 1.0 - How can i do that?

[TomDoan]

Use the VTICKS option on GRAPH. If the numbers are between 0 and 1, VTICKS=11 will probably give you the 0, .1, ... The default for VTICKS=9, so it wouldn't be able to label a [0,1] range by .1, so it does .2 instead. Note that with IRF's, you generally don't want to use even more ticks because the axis gets crowded when you have more than one graph on a page.