system is computationally singular: reciprocal number condition

Asked

Viewed 1,010 times

2

I’m dealing with time series on R. My Data are Homocedastic, have no autocorrelation, no unitary root in lag=0, trend or seasonality.

I still run into the problem of this particular error:

    VAR_S <- summary(VAR(Base_Dados,p = 1, type = "both"))

Error in Solve.default(Sigma) :

system is computationally singular: reciprocal condition number =
5.67068e-37

VAR<- (VAR(Base_Dados, p = 1, type = "both" ) )

At first I thought it was collinearity, but the correlation of the data does not exceed 0,25.

r

  • You may be working with something really independent, so the autoregressive vector tends to a null vector.

  • I think I understand, is it possible to see this with determinants of the matrix maybe?

1 answer

1

The matrix cannot be reversible the determinant of the matrix is zero or approximately zero. I have dealt with this in two different ways, you have changed the data to a similar model that has worked poorly. as for the previous model the form found of resolution, was to re evaluate the scale of the variables they ceased to be in the same unit scale but began to have approximately the same amount of integer number per observation.

It was intuitive work, so I have no formulas to demonstrate.

Browser other questions tagged r

You are not signed in. Login or sign up in order to post.