ABSTRACT
Continuing on the theme of the last chapter, in this chapter we introduce three other order determination methods. The first is a class of BIC-type criteria that maximizes the information while penalizing the number of parameters. Similar to the sequential test, this method relies on the magnitude of the eigenvalues. The second method, in contrast, relies on the bootstrapped eigenvector variation, which tends to be larger in the null space of the candidate matrix than in the range space of the matrix. The third method combines the advantages of the first two methods by exploiting both the magnitude of eigenvalues and the bootstrapped eigenvector variations.
