VAR Order Selection

[Nabtheberliner]

Hello Tom,
I'm on the CH IV " New Introduction To Multiple Time Series analysis" I'm dealing with the Numerical Problems p.191
Problem 4.6 based on the file E2

Set up a sequence of tests for the correct VAR order of the data generating
process using a maximum order of M = 4. Compute the required χ2 and F
likelihood ratio statistics. Which order would you choose?

which is related to section 4.2.4.

To answer the question i write the program:

Code: Select all

OPEN DATA "C:\Users\amaraoui nabih\Desktop\ECONOMIE\MANUELS\NEW INTRODUCTION TO MULTIPLE TIME SERIES ANALYSIS\Problems CH IV p191\E2.txt"
CALENDAR(Q) 1947:1
DATA(FORMAT=PRN,ORG=COLUMNS) 1947:01 1972:04 y1 y2

@varlagselect(crit=gtos,det=constant,lags=4) * 1972:4
# y1 y2

The output is:

Code: Select all

VAR Lag Selection
Lags F-Test
   0   0.000000*
   1 526.013725
   2  45.768352
   3   4.783891
   4   6.805817

Now i check with the table 4.4 p.146 based on file westgermaneco.txt (E1 for Lütkepohl's book) and what i don't get is that i have F-Test with what seems to be LR Statistics ~khi2 and not ~F distribution
So i checked the appropiate program from Lütkepohl which is related to this table 4.4:


Lutkepohl, New Introduction, example 4.2.4 from pp 145-146
* VAR; Lag selection

Code: Select all

open data e1.dat
calendar(q) 1960
data(format=prn,org=columns,skips=6) 1960:01 1982:04 invest income cons
*
set dinc  = log(income/income{1})
set dcons = log(cons/cons{1})
set dinv  = log(invest/invest{1})
*
@varlagselect(crit=gtos,det=constant,lags=4) * 1978:4
# dinc dcons dinv

and the output is:

Code: Select all

VAR Lag Selection
Lags F-Test
   0  0.0000000*
   1 23.2488409
   2 24.9009035
   3  4.7563989
   4 14.4383478

which doesn't really fit to the results in the table. I don't know how to read properly this.
The column that supposed to follow a khi2 distribution up here follows a F-Test?
Do you see what i mean?
And how does he get the critical values?

My second question still about Order Selection is page 148:
i don't find the same results even with the Lütkepohl program

Code: Select all

*
* Lutkepohl, New Introduction, example 4.3 from p. 148
* VAR; Lag selection
*
open data e1.dat
calendar(q) 1960
data(format=prn,org=columns,skips=6) 1960:01 1982:04 invest income cons
*
set dinc  = log(income/income{1})
set dcons = log(cons/cons{1})
set dinv  = log(invest/invest{1})
*
@varlagselect(lags=4,det=constant,crit=aic) * 1978:4
# dinc dcons dinv
@varlagselect(lags=4,det=constant,crit=sbc) * 1978:4
# dinc dcons dinv
@varlagselect(lags=4,det=constant,crit=hq)  * 1978:4
# dinc dcons dinv

The output is

Code: Select all

VAR Lag Selection
Lags AICC
   0 -1123.4537
   1 -1127.2573
   2 -1130.8805*
   3 -1112.2548
   4 -1100.8783


VAR Lag Selection
Lags SBC/BIC
   0 -1116.7804*
   1 -1101.6652
   2 -1088.2020
   3 -1054.5942
   4 -1030.6685


VAR Lag Selection
Lags HQ
   0 -1120.8691*
   1 -1118.0198
   2 -1116.8225
   3 -1095.4807
   4 -1083.8209

Finally, how do you get FPE Criterion?

I hope my questions are clear enough.
Thanks a lot Tom

[Nabtheberliner]

i think the E2 file wasn't attached, here it is

[TomDoan]

There's a newer version of @VARLagSelect posted at

http://www.estima.com/forum/viewtopic.php?f=7&t=1261

That corrects the labeling on the GTOS choice, which is strictly done as a LR test with a chi-squared asymptotic distribution, not an F.

[Nabtheberliner]

I check this out. I gess my second question is linked to the first one and find the answer with this procedure or the question isn't relevant?
Thanks for your help Tom