3:58 a.m.
Hello,
For my school these, i am doing research to investigate the effect of
different variables on the reserve currency shares of different currencies.
Before i am working with panel data, i liked to make an OLS estimation on
the dollar. So i'm using time series data. I checked the variables for
stationarity using the ADF-test, and it seems that the dependent variable
is non-stationary along with two explanatory variables. The other 4
explanatory variables are stationary. Since you cannot run regressions with
variables of a different integration, i am not sure how to proceed.. I
suggest it is not appropriate to just take the first differences of the
non-stationary variables?
Any help on this topic would be very much appreciated. Thanks in advance.
For additional info on the data/research, please do not hesitate to contact
me.
--
Met vriendelijke groeten,
Emiel Denuwelaere
M: +32 (0)476 56 08 75
E: emiel.denuwelaere(a)gmail.com
4:04 a.m.
Emiel
As far as I know, you can run a regression with regressors that are
originally of different integration orders, provided you first
difference the I(1) variables, so that your regressions contains only
I(0) variables. This is done and advocated for example (although
slightly different method) in Juselius book on cointegrated VAR, where
some variables are kept I(0).
Best
Matthieu
Le 10/08/2012 09:58, Emiel Denuwelaere a écrit :
Hello,
For my school these, i am doing research to investigate the effect of
different variables on the reserve currency shares of different
currencies. Before i am working with panel data, i liked to make an
OLS estimation on the dollar. So i'm using time series data. I checked
the variables for stationarity using the ADF-test, and it seems that
the dependent variable is non-stationary along with two explanatory
variables. The other 4 explanatory variables are stationary. Since you
cannot run regressions with variables of a different integration, i am
not sure how to proceed.. I suggest it is not appropriate to just take
the first differences of the non-stationary variables?
Any help on this topic would be very much appreciated. Thanks in advance.
For additional info on the data/research, please do not hesitate to
contact me.
--
Met vriendelijke groeten,
Emiel Denuwelaere
M: +32 (0)476 56 08 75
E: emiel.denuwelaere(a)gmail.com <mailto:emiel.denuwelaere@gmail.com>
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
4:50 a.m.
Hi,
allow me a meta remark first: This is not a gretl-specific question, and
IMHO does not really belong on this list.
Having said that, from the description of the problem the mentioned
"solution" is certainly not what Katarina Juselius would suggest. If you
have several I(1) variables, fine, look for cointegration among them.
Nothing prevents you from adding further I(0) variables to your model
afterwards. However, note the "if" part: since you seem to be talking
about shares, the non-stationarity in these variables should be
something other than I(1)-ness, at least strictly speaking. There is no
general rule how to proceed, opinions differ.
cheers,
sven
On 08/10/2012 10:04 AM, Matthieu Stigler wrote:
Emiel
As far as I know, you can run a regression with regressors that are
originally of different integration orders, provided you first
difference the I(1) variables, so that your regressions contains only
I(0) variables. This is done and advocated for example (although
slightly different method) in Juselius book on cointegrated VAR, where
some variables are kept I(0).
Best
Matthieu
Le 10/08/2012 09:58, Emiel Denuwelaere a écrit :
> Hello,
>
> For my school these, i am doing research to investigate the effect of
> different variables on the reserve currency shares of different
> currencies. Before i am working with panel data, i liked to make an
> OLS estimation on the dollar. So i'm using time series data. I checked
> the variables for stationarity using the ADF-test, and it seems that
> the dependent variable is non-stationary along with two explanatory
> variables. The other 4 explanatory variables are stationary. Since you
> cannot run regressions with variables of a different integration, i am
> not sure how to proceed.. I suggest it is not appropriate to just take
> the first differences of the non-stationary variables?
>
> Any help on this topic would be very much appreciated. Thanks in advance.
> For additional info on the data/research, please do not hesitate to
> contact me.
>
> --
> Met vriendelijke groeten,
>
> Emiel Denuwelaere
>
> M: +32 (0)476 56 08 75
> E: emiel.denuwelaere(a)gmail.com <mailto:emiel.denuwelaere@gmail.com>
>
>
>
> _______________________________________________
> Gretl-users mailing list
> Gretl-users(a)lists.wfu.edu
> http://lists.wfu.edu/mailman/listinfo/gretl-users
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
5:58 a.m.
Dear Emiel
This is interesting. But the logic is very clear , If there is no
relationship between two variables , why we should try to establish one.
I am not sure but cannot a factor analysis give some guidance as to reduce
the concerned variables
I also think some omitted variable may do the trick
I feel these suggestions may not exacrly address the problem in hand then
soory to give suggestions.
N.K.Dashora
Udaipur ( Rajasthan )
India
On Fri, Aug 10, 2012 at 1:28 PM, Emiel Denuwelaere <
emiel.denuwelaere(a)gmail.com> wrote:
Hello,
For my school these, i am doing research to investigate the effect of
different variables on the reserve currency shares of different currencies.
Before i am working with panel data, i liked to make an OLS estimation on
the dollar. So i'm using time series data. I checked the variables for
stationarity using the ADF-test, and it seems that the dependent variable
is non-stationary along with two explanatory variables. The other 4
explanatory variables are stationary. Since you cannot run regressions with
variables of a different integration, i am not sure how to proceed.. I
suggest it is not appropriate to just take the first differences of the
non-stationary variables?
Any help on this topic would be very much appreciated. Thanks in advance.
For additional info on the data/research, please do not hesitate to
contact me.
--
Met vriendelijke groeten,
Emiel Denuwelaere
M: +32 (0)476 56 08 75
E: emiel.denuwelaere(a)gmail.com
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
7 p.m.
New subject: MacKinnon critical values
Hi all.
I'm testing a time series with the ADF unit root test and according to
Gretl I have to conmpare the value of the tau-statistic with
MacKinnon's critical values (to reject/accept the null of the presence
of a unit root)...
The question is:
do I have to tabulate them? Is there any way I can do it through a
Gretl command?
"beta0 + beta1/T + beta2/T^2 + beta3/T^3"...could this be a correct
way to compute it directly on Gretl?
Thank you all.
----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.
**** Riservatezza / Confidentiality ****
In ottemperanza al D.Lgs. n. 196 del 30/6/2003 in materia di protezione dei dati
personali, le informazioni contenute in questo messaggio sono strettamente riservate ed
esclusivamente indirizzate al destinatario indicato (oppure alla persona responsabile di
rimetterlo al destinatario). Vogliate tener presente che qualsiasi uso, riproduzione o
divulgazione di questo messaggio e' vietato. Nel caso in cui aveste ricevuto questo
messaggio per errore, vogliate cortesemente avvertire il mittente e distruggere il
presente messaggio.
8:17 p.m.
New subject: MacKinnon critical values
On Tue, 14 Aug 2012, Alessia Via wrote:
I'm testing a time series with the ADF unit root test and
according to
Gretl I have to conmpare the value of the tau-statistic with
MacKinnon's critical values (to reject/accept the null of the presence
of a unit root)...
Gretl prints out the p-value based on MacKinnon's approximation to
the distribution of these statistics. You don't have to do anything,
just know what a p-value is.
Allin Cottrell
3:20 a.m.
New subject: MacKinnon critical values
Thank you Allin!
That means that if I have something like this:
tau_c(1) = -2,69871
asymptotic p-value : 0,07419
I can simply not reject the null
while:
tau_c(1) = -4,15522
asymptotic p-value : 0,0007807
means I can reject the null? Without further computations?
Def. Quota Allin Cottrell <cottrell(a)wfu.edu>:
On Tue, 14 Aug 2012, Alessia Via wrote:
> I'm testing a time series with the ADF unit root test and according to
> Gretl I have to conmpare the value of the tau-statistic with
> MacKinnon's critical values (to reject/accept the null of the presence
> of a unit root)...
Gretl prints out the p-value based on MacKinnon's approximation to
the distribution of these statistics. You don't have to do anything,
just know what a p-value is.
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.
**** Riservatezza / Confidentiality ****
In ottemperanza al D.Lgs. n. 196 del 30/6/2003 in materia di protezione dei dati
personali, le informazioni contenute in questo messaggio sono strettamente riservate ed
esclusivamente indirizzate al destinatario indicato (oppure alla persona responsabile di
rimetterlo al destinatario). Vogliate tener presente che qualsiasi uso, riproduzione o
divulgazione di questo messaggio e' vietato. Nel caso in cui aveste ricevuto questo
messaggio per errore, vogliate cortesemente avvertire il mittente e distruggere il
presente messaggio.
4:33 a.m.
New subject: MacKinnon critical values
Am 14.08.2012 09:20, schrieb Alessia Via:
Thank you Allin! That means that if I have something like this:
tau_c(1) = -2,69871 asymptotic p-value : 0,07419
I can simply not reject the null
while:
tau_c(1) = -4,15522 asymptotic p-value : 0,0007807
means I can reject the null? Without further computations?
Hi, I can see two possible interpretations of your question:
Either you're asking what a p-value is. This would (a) not be
gretl-specific and (b) very basic statistics, such that honestly I would
recommend not to do any empirical work on your own.
Or you're asking whether gretl prints the appropriate p-values for this
test. Allin gave the answer already: yes they are (apart from advanced
details such as small-sample issues etc., but that's why gretl prints
the modifier "asymptotic").
cheers,
sven
4:52 a.m.
New subject: MacKinnon critical values
Hi Sven,
It is all very clear (p-value too!Thank you) but, when I tabulated the
critical values using
"beta0 + beta1/T + beta2/T^2 + beta3/T^3"
there were cases in which there were, in my opinion, conflicting
results (even if the differences were very very small).That is why I
was wondering if there was the need to further verify....
In any case, the be-all and end-all is research and study and...I
would encourage it.
Cheers
Def. Quota Sven Schreiber <svetosch(a)gmx.net>:
Am 14.08.2012 09:20, schrieb Alessia Via:
> Thank you Allin! That means that if I have something like this:
>
> tau_c(1) = -2,69871 asymptotic p-value : 0,07419
>
> I can simply not reject the null
>
> while:
>
> tau_c(1) = -4,15522 asymptotic p-value : 0,0007807
>
> means I can reject the null? Without further computations?
>
Hi, I can see two possible interpretations of your question:
Either you're asking what a p-value is. This would (a) not be
gretl-specific and (b) very basic statistics, such that honestly I would
recommend not to do any empirical work on your own.
Or you're asking whether gretl prints the appropriate p-values for this
test. Allin gave the answer already: yes they are (apart from advanced
details such as small-sample issues etc., but that's why gretl prints
the modifier "asymptotic").
cheers,
sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.
**** Riservatezza / Confidentiality ****
In ottemperanza al D.Lgs. n. 196 del 30/6/2003 in materia di protezione dei dati
personali, le informazioni contenute in questo messaggio sono strettamente riservate ed
esclusivamente indirizzate al destinatario indicato (oppure alla persona responsabile di
rimetterlo al destinatario). Vogliate tener presente che qualsiasi uso, riproduzione o
divulgazione di questo messaggio e' vietato. Nel caso in cui aveste ricevuto questo
messaggio per errore, vogliate cortesemente avvertire il mittente e distruggere il
presente messaggio.
4:56 a.m.
New subject: MacKinnon critical values
Am 14.08.2012 10:52, schrieb Alessia Via:
Hi Sven,
It is all very clear (p-value too!Thank you) but, when I tabulated
the critical values using "beta0 + beta1/T + beta2/T^2 + beta3/T^3"
there were cases in which there were, in my opinion, conflicting
results (even if the differences were very very small).That is why I
was wondering if there was the need to further verify....
So you're trying to replicate MacKinnon's results, is that it? If you
think you've found differences with respect to gretl's implementation,
then you would need to show us those differences. But be aware that this
is a pretty well-tested area in gretl I think, so maybe you should check
your results first.
-s
5:19 a.m.
New subject: MacKinnon critical values
Sorry for the misunderstanding...Of course, I think my results are
wrong (and need further checks)
:)
Def. Quota Sven Schreiber <svetosch(a)gmx.net>:
Am 14.08.2012 10:52, schrieb Alessia Via:
> Hi Sven,
>
> It is all very clear (p-value too!Thank you) but, when I tabulated
> the critical values using "beta0 + beta1/T + beta2/T^2 + beta3/T^3"
> there were cases in which there were, in my opinion, conflicting
> results (even if the differences were very very small).That is why I
> was wondering if there was the need to further verify....
>
So you're trying to replicate MacKinnon's results, is that it? If you
think you've found differences with respect to gretl's implementation,
then you would need to show us those differences. But be aware that this
is a pretty well-tested area in gretl I think, so maybe you should check
your results first.
-s
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.
**** Riservatezza / Confidentiality ****
In ottemperanza al D.Lgs. n. 196 del 30/6/2003 in materia di protezione dei dati
personali, le informazioni contenute in questo messaggio sono strettamente riservate ed
esclusivamente indirizzate al destinatario indicato (oppure alla persona responsabile di
rimetterlo al destinatario). Vogliate tener presente che qualsiasi uso, riproduzione o
divulgazione di questo messaggio e' vietato. Nel caso in cui aveste ricevuto questo
messaggio per errore, vogliate cortesemente avvertire il mittente e distruggere il
presente messaggio.
4296
days inactive
4300
days old
10 comments
6 participants
participants (6)
-
Alessia Via -
Allin Cottrell -
Emiel Denuwelaere -
Matthieu Stigler -
Narandra Dashora -
Sven Schreiber