Page 1 of 1
Mishkin Test
Posted: Mon Jan 31, 2011 1:08 am
by abbas
Dears
I want to run the Mishkin (1983) test with Eviews 7. This needs a program with programming tools in Eviews. Can you help me?
Thanks
Abbas Aflatooni
E-mail:
abbasaflatooni@gmail.com
Re: Mishkin Test
Posted: Mon Jan 31, 2011 10:30 am
by EViews Gareth
My very quick look seems to indicate that it is simply a case of estimating a system with two equations in it, then doing a Wald test (the original paper says an LR test, but a Wald test will do) on whether alpha1 = alpha1*. Something like this (with generated data):
Code: Select all
create u 100
series ret=nrnd
series earn=nrnd
rnd(c) 'generate random starting values
system s
s.append earn=c(1)+c(2)*earn(-1)
s.append ret = c(3)*(earn-c(1)-c(4)*earn(-1))
s.ls 'estimate system
s.wald c(2)=c(4) 'do Wald test of a1=a1*
Re: Mishkin Test
Posted: Tue Feb 01, 2011 2:11 am
by abbas
Dears,
At firs I think that the Wald test is correct but it is not a correct respones. I wrote the following program, now (LR test). is it work correctly, please?
Thank you
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
create u 100
series ret=nrnd
series earn=nrnd
rnd(c) 'generate random starting values
equation regobs.ls returns c earnings(-1)
scalar n=@regobs
system mysys1
mysys1.append earnings=c(1)+c(2)*earnings(-1)
mysys1.append returns=c(3)*(earnings-c(4)-c(5)*earnings(-1))
mysys1.sur
sym ssru=mysys1.@ssr(1)*mysys1.@ssr(2)
system mysys2
mysys2.append earnings=c(1)+c(2)*earnings(-1)
mysys2.append returns=c(3)*(earnings-c(4)-c(2)*earnings(-1))
mysys2.sur
sym ssrc1=mysys2.@ssr(1)*mysys2.@ssr(2)
scalar Mishkin1 = 2*n*log(@det(ssrc1)/@det(ssru))
table(4,4) result
setcell(result,1,1, "Mishkin Test by Aflatooni")
setline(result,2)
setcell(result,3,1,"Mishkin1=")
setcell(result,3,2,Mishkin1)
setcell(result,3,3,"p-value = ")
setcell(result,3,4,1-@cchisq(Mishkin1,1))
setline(result,4)
show result
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
Re: Mishkin Test
Posted: Tue Feb 01, 2011 2:13 am
by abbas
Dears,
At firs I think that the Wald test is correct but it is not a correct respones. I wrote the following program, now (LR test). is it work correctly, please? (Final Version)
Thank you
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
create u 100
series returns=nrnd
series earnings=nrnd
rnd(c) 'generate random starting values
equation regobs.ls returns c earnings(-1)
scalar n=@regobs
system mysys1
mysys1.append earnings=c(1)+c(2)*earnings(-1)
mysys1.append returns=c(3)*(earnings-c(4)-c(5)*earnings(-1))
mysys1.sur
sym ssru=mysys1.@ssr(1)*mysys1.@ssr(2)
system mysys2
mysys2.append earnings=c(1)+c(2)*earnings(-1)
mysys2.append returns=c(3)*(earnings-c(4)-c(2)*earnings(-1))
mysys2.sur
sym ssrc1=mysys2.@ssr(1)*mysys2.@ssr(2)
scalar Mishkin1 = 2*n*log(@det(ssrc1)/@det(ssru))
table(4,4) result
setcell(result,1,1, "Mishkin Test by Aflatooni")
setline(result,2)
setcell(result,3,1,"Mishkin1=")
setcell(result,3,2,Mishkin1)
setcell(result,3,3,"p-value = ")
setcell(result,3,4,1-@cchisq(Mishkin1,1))
setline(result,4)
show result
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
Mishkin Test
Posted: Tue Feb 01, 2011 7:57 am
by EViews Gareth
Methodology seems about right, although I think you might want to restrict the constant to be the same in the two equations, rather than letting them differ.
I'm also not sure you're calculating the SSRs correctly.