[R-SIG-Finance] Fwd: Re: Block Exogeneity Test

John Frain frainj at tcd.ie
Sun Nov 23 13:31:59 CET 2014


Debasish


Sorry if I misunderstood your example. I have seen many economists
trying to estimate a system similar to what you set out. The problems
are basic to any VAR analysis and are not simply "operational nuances
to deal with non-stationarity".

Apart for this I am not sure that I understand what you are trying to
do. My understanding of block exogeneity is as follows. If Y is an n x
1 vector of (stationary) variables is partitioned into X (m x 1) (the
first m variables in Y) and Z(k X 1) (n  = m+k), Then the matrix of
lag polynomials in the VAR for  Y separates into 4 blocks with the
matrix of lag polynomials of the X VAR in the upper left hand corner
and that of Y in the lower right corner and zero polynomials
elsewhere.  A test for this block exogeneity is in effect a test for
that the  zero polynomials are zero. I have not encountered such a
test in R. However it should be easy to implement as a likelihood
ratio test either jointly or by variable if this is what you want to
do.  (While I remember seeing block exogeneity mentioned in the
literature I am not sure of their significance. I can not quote any
references.)

You have stated elsewhere that "I was looking for the effect of lags
of one variable on the other by excluding the lags of other
variables". Is that not just a case of estimating a general model and
then testing the validity of the implied restrictions" using LR, ML or
Wald type tests.

Best Regards

John


John C Frain, Ph.D.
Economics Department             3 Aranleigh Park
Trinity College Dublin                 Rathfarnham
College Green                           Dublin 14
Dublin 2                                    Ireland
Ireland
www.tcd.ie/Economics/staff/frainj/home.htm
mailto:frainj at tcd.ie
mailto:frainj at gmail.com


On 22 November 2014 at 07:56, DEBASISH MAITRA <debasishmaitra at gmail.com> wrote:
> Dear John,
>
> I know that these variables are definitely non-stationary, even that I
> mentioned in the bracket. But I was only looking for block exogeneity wald
> test, not operational nuances to deal with non-stationarity. I was only
> interested to know whether we have command under vars package to know this,
> because MSBVAR and VARS both give only bi-variate causality.
>
> Thanks!
> Regards,
> Debasish
>
> On Sat, Nov 22, 2014 at 3:25 AM, John Frain <frainj at tcd.ie> wrote:
>>
>> This kind of econometrics gives econometricians a bad name. The
>> variables gdp, cpi and m3 are certainly not stationary. Modelling
>> should be in the logarithms of the variables and not in levels. If
>> there is no cointegration and the logs of the three variables are I(1)
>> VAR modelling should be in first differences of the logs.
>>
>> The simplest way to test for block exogeneity would be (dlgdp is the
>> first difference of the log of GDP)
>>
>> 1. Estimate, by OLS, an AR(p) process for dlgdp and extract the
>> loglikelihood
>> 2. Estimate by OLS the dlgdp equation from the VAR(p) involving dlgdp
>> dlcpi dlm3 and extract the loglikelihood.
>> 3. Use the loglikelihood ratio test to test the significance of the
>> extra variables in 2.  This is a block exogeneity test of dlcpi and
>> dlm3 for dlgdp.
>>
>> Working in this way ensures that one understands what they are doing
>> and adds to your understanding of the process..
>>
>> If there is cointegration you will have to think a little more.
>> Lutkepohl, Helmut  (2005), New Introduction to Multiple Time Series,
>> Springer is one of the many good references available on this topic
>>
>> Best Regards
>>
>> John
>>
>> John C Frain, Ph.D.
>>
>> Economics Department             3 Aranleigh Park
>> Trinity College Dublin                 Rathfarnham
>> College Green                           Dublin 14
>> Dublin 2                                    Ireland
>> Ireland
>> www.tcd.ie/Economics/staff/frainj/home.htm
>> mailto:frainj at tcd.ie
>> mailto:frainj at gmail.com
>>
>>
>> On 21 November 2014 11:03, Brian G. Peterson <brian at braverock.com> wrote:
>> > For the Wald-type test, see
>> >
>> > ###
>> >
>> > ?causality
>> >
>> > ###
>> >
>> > the vars package calls this 'instantaneous causality'.
>> >
>> > You will need to set the 'cause' parameter to a vector of your
>> > predictors.
>> > In your example:
>> >
>> > ###
>> >
>> > cause=c("gdp","cpi","m3")
>> >
>> > # note that R is case sensitive, and your
>> > # non-reproducible example mixed case
>> >
>> > ###
>> >
>> > You likely also want boot=TRUE and boot.runs set to some value
>> > appropriate
>> > to your data, in most formulations of the test procedure you're seeking.
>> >
>> > As you've described it, you're looking for bivariate tests, not truly a
>> > block exogeneity test, which tests whether the group of variables or
>> > 'block'
>> > are tested to see if the *block* causes any of the observed lags.  It
>> > appears from the output that you request that you are looking for
>> > pairwise
>> > causality as individual variables are tested against the rest of the
>> > model.
>> >
>> > Regards,
>> >
>> > Brian
>> >
>> >
>> > On 11/21/2014 04:49 AM, Brian G. Peterson wrote:
>> >>
>> >> To the list. - Brian
>> >>
>> >>
>> >> -------- Forwarded Message --------
>> >> Subject:     Re: [R-SIG-Finance] Block Exogeneity Test
>> >> Date:     Fri, 21 Nov 2014 16:05:44 +0530
>> >> From:     DEBASISH MAITRA <debasishmaitra at gmail.com>
>> >> To:     Brian G. Peterson <brian at braverock.com>
>> >>
>> >>
>> >>
>> >> Hi Brian,
>> >>
>> >> Thanks for your mail.
>> >> I am specifically looking for Block Exogeneity Wald test after my VAR
>> >> model. For an example, I have three series gdp, cpi and m3. After
>> >> estimating the VAR model (with an assumption that these three variable
>> >> are stationary) among these three varables, if I really want to test
>> >> hypotheses ; lags of cpi explain current gdp, lags of cpi explain
>> >> current m3; lags of gdp explain current cpi, lags of gdp explain
>> >> current
>> >> m3 etc. I should have the results look like;
>> >>
>> >> Dependent variable GDP
>> >> Excluded                Chi-sq                df               prob
>> >> cpi
>> >> m3
>> >>
>> >> Dependent variable cpi
>> >> Excluded               Chi-sq                 df               prob
>> >> gdp
>> >> m3
>> >>
>> >> Dependent variable m3
>> >> Excluded               Chi-sq                 df               prob
>> >> gdp
>> >> cpi
>> >>
>> >> Kindly look into this and suggest the way ou.
>> >>
>> >>
>> >> Regards,
>> >> Debasish
>> >>
>> >> On Fri, Nov 21, 2014 at 3:23 PM, Brian G. Peterson <brian at braverock.com
>> >> <mailto:brian at braverock.com>> wrote:
>> >>
>> >>      On 11/21/2014 03:31 AM, DEBASISH MAITRA wrote:
>> >>
>> >>          I am Debasish Maitra. I am trying to run block exogeneity test
>> >>          in R (vars
>> >>          package) after a VAR model, but I am not able to find anything
>> >>          out to do
>> >>          this.
>> >>
>> >>          It would be of great help if you kindly tell me how can I run
>> >> this.
>> >>
>> >>
>> >>      It would be a great help if you followed the posting guide:
>> >>
>> >>      http://www.r-project.org/__posting-guide.html
>> >>      <http://www.r-project.org/posting-guide.html>
>> >>
>> >>      and provided a reproducible example.  Use a data set included with
>> >>      the vars package, or some other data set included with a different
>> >> R
>> >>      package, and tell us in a reproducible way what you've already
>> >>      tried, and where you're stuck.
>> >>
>> >>      I don't believe that the 'vars' package includes a function for a
>> >>      block exogeneity test on its model output.
>> >>
>> >>      The  MSBVAR package provides bivariate Granger causality tests in
>> >>      lieu of a block exogeneity test.
>> >>
>> >>      It would also help if you would be explicit about which form of
>> >>      exogeneity test you are trying to apply, with a reference to the
>> >>      paper you hope to use the technique from.  I'm aware of at least
>> >> two
>> >>      different approaches, which each have different assumptions.
>> >>
>> >>      Regards,
>> >>
>> >>      Brian
>> >>
>> >>      --
>> >>      Brian G. Peterson
>> >>      http://braverock.com/brian/
>> >>      Ph: 773-459-4973 <tel:773-459-4973>
>> >>      IM: bgpbraverock
>> >>
>> >>      _________________________________________________
>> >>      R-SIG-Finance at r-project.org <mailto:R-SIG-Finance at r-project.org>
>> >>      mailing list
>> >>      https://stat.ethz.ch/mailman/__listinfo/r-sig-finance
>> >>      <https://stat.ethz.ch/mailman/listinfo/r-sig-finance>
>> >>      -- Subscriber-posting only. If you want to post, subscribe first.
>> >>      -- Also note that this is not the r-help list where general R
>> >>      questions should go.
>> >>
>> >>
>> >>
>> >>
>> >> --
>> >>
>> >> _______________________________________________
>> >> R-SIG-Finance at r-project.org mailing list
>> >> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
>> >> -- Subscriber-posting only. If you want to post, subscribe first.
>> >> -- Also note that this is not the r-help list where general R questions
>> >> should go.
>> >
>> >
>> >
>> > --
>> > Brian G. Peterson
>> > http://braverock.com/brian/
>> > Ph: 773-459-4973
>> > IM: bgpbraverock
>> >
>> >
>> > _______________________________________________
>> > R-SIG-Finance at r-project.org mailing list
>> > https://stat.ethz.ch/mailman/listinfo/r-sig-finance
>> > -- Subscriber-posting only. If you want to post, subscribe first.
>> > -- Also note that this is not the r-help list where general R questions
>> > should go.
>>
>> _______________________________________________
>> R-SIG-Finance at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
>> -- Subscriber-posting only. If you want to post, subscribe first.
>> -- Also note that this is not the r-help list where general R questions
>> should go.
>
>
>
>
>



More information about the R-SIG-Finance mailing list