Balcilar, Gupta and Miller, EE 2015

[LvXin]

Dear Tom,
I also could not find which parameters indicate the time duration of regime 1 and regime 2. The time duration means how long will regime 1 or regime 2 will be persistent.

Best Regards!

[TomDoan]

Expected duration is 1/(prob remaining in regime).

[LvXin]

Yes! Thanks very much. I nearly forget this equition! I could caculate it by myself!

Best Regard!

[bekkdcc]

Dear Tom,

I am replicating Balcilar, Gupta and Miller, EE 2015 with their data set and your program (Fuzz" up th............). I can not able to get some values and can not find how to get them.

1. stock-warson cointegration test

2. the prob value of regime properties in Table 4 which are( 0.72 and 0.27)

3. Davies test

4. high - lov volatilty and their count in table 5.


I will be so pleased if you helped me on this issues, cause after I got this informations I am going to do some changes in code for my study, thats why I should to learn these.

[TomDoan]

Dear Tom,

I am replicating Balcilar, Gupta and Miller, EE 2015 with their data set and your program (Fuzz" up th............). I can not able to get some values and can not find how to get them.

1. stock-warson cointegration test

That's done with the @SWTRENDS procedure. See the swtrends.src file for a description. That isn't used very much.

3. Davies test

See https://estima.com/forum/viewtopic.php?p=15185. The use of the Davies adjustment for MS models dates back to an earlier paper which summarizes it incorrectly.

2. the prob value of regime properties in Table 4 which are( 0.72 and 0.27)
4. high - lov volatilty and their count in table 5.

The "long-run average" probabilities is another name for the ergodic probabilities. %mcergodic(p) will compute those. As to the "actual outcomes", I assume they counted the number of observations where the probability of a particular regime was greater than .5. Since the actual regime is never observable, that's not a particularly meaningful result. The sum of the smoothed probabilities (that is, if the probability of regime 1 is .6, then count regime 1 as .6 and regime 2 as .4) will generally be almost exactly equal to those long-run probabilities. The "actual outcomes" would count the .6 as 1 for regime 1 and 0 for regime 2, even though there is considerable doubt as to which regime actually holds.

[bekkdcc]

Dear Tom,

1. I found the function whch is below %mcergodic but it didn not give any number, what shoukd I write at the end of the code to get long-run average" probabilities or ergodic probabilities as a number(value)


function %mcergodic p
type vect %mcergodic
type rect p
*
local rect a
local integer i j n
*
compute n=%cols(p)
dim a(n+1,n)
ewise a(i,j)=%if(i>n,1,(i==j)-%if(i<n,p(i,j),1-%sum(%xcol(p,j))))
compute %mcergodic=%xcol(inv(%innerxx(a))*tr(a),n+1)
end

2. I also want to get loadings in the @johmle, so what should I do

[TomDoan]

Dear Tom,

1. I found the function whch is below %mcergodic but it didn not give any number, what shoukd I write at the end of the code to get long-run average" probabilities or ergodic probabilities as a number(value)

Don't change the function. It produces a VECTOR. Just DISPLAY it.

display %mcergodic(p)

2. I also want to get loadings in the @johmle, so what should I do

Use the LOADINGS option on @JOHMLE.