Determine lag length in VAR model

[nesreen]

Dear All,

I am running a 4 variable VAR using monthly data (117 observations) . In determining the lag length to be used in the model:
1- I started with a maximum lag length equals 12,
2- Run VAR in levels for each lag; starting with lag 1 12 and moving backwards 1 11 and so on.
3- Find Akaiake (AIC) and Schwarz (SIC) for VAR model for each lag length and choose the lag length that minimizes AIC and SIC.

I have 2 questions:

1- I found that AIC and SIC are negative, in this case should I choose the lag length based on minimum absolute value of information criteria or should I choose lag length based on minimum value for information criteria taking sign into consideration?

2- AIC and SIC gives conflicting results:
At lag length 1 12 AIC= -6.178598 and SIC= -1.224539
At lag length 1 6 AIC= -4.691615 and SIC= -2.250597
At lag length 1 1 AIC= -4.096655 and SIC= -3.621899
if i choose lag length based on minimum absolute value: AIC implies lag length 1 1 , while SIC implies lag length 1 12
if i choose lag length based on minimum value taking sign into consideration: AIC implies lag length 1 12 , while SIC implies lag length 1 1
which one should I base my decision on?

I would really appreciate your response as soon as possible,

Thank you very much,
Nesreen