Proc VARMAX

[Fery123]

I find a problem when I want to model
vector autoregressive exogenous, instantly the message "The lag 0 covariance matrix is ​​singular." and The model is not full rank. The VARMAX procedure halts further steps

data finance;
input y1 y2 y3 y4;
cards;
2.22 3.21 2.22 1.43
1.32 2.45 2.31 2.32
2.22 3.21 2.22 1.43
1.32 2.45 2.31 2.32
2.22 3.21 2.22 1.43
:
1.32 2.81 2.31 2.56
2.82 3.21 2.22 1.43
;
run;
proc varmax data = finance;
model y1 y2 y3 = y4 / p = 1 xlag = 1 nocurrentx noint;
run;

[stat_sas]

Hi,

Varmax's parameter estimation is based on least squares which requires inverse of a matrix in its calculation. If some of the variables are highly or perfectly correlated then matrix will not invert and you will get this error.