• • • <lk ≤ n. Note that
i ≤ li , for all i ≤ k.
Therefore
max(i, j) ≤ max(li, lj), for i, j ≤ k.
If equality does not hold for some i and j, then by (1) we have,
aij ≥ ali,lj.
If equality does hold for some i and j, then let lt = max(li, lj ). Note that
i < j ⇔ li < lj . So we have t = max(i, j) and therefore lt = t. Hence
lt-1 = t - 1, . . . , l1 = 1. In this case
aij = ali,lj.
Either way, aij ≥ ali ,lj holds. Therefore
m(A(l1, . . . lk)) ≤ m(A(1, . . . k))
= m(A[k])
= δk(A),
as A(l1 , . . . lk) was arbitrary. Hence the result.
Example 14. Consider the matrix
9 91 |
8 |
4 |
3 ʌ | ||
A = |
8 |
6 |
5 |
4 | |
5 |
4 |
4 |
3 | ||
3 |
2 |
3 |
1 |
/ |
The indicated lines help to check that A is pyramidal. Hence by Theorem 13
we find:
δ1(A) = m(A[1]) = 9
δ2 (A) = m(A[2]) = 16
δ3 (A) = m(A[3]) = 20
δ4(A) = m(A[4]) = m(A) = 22.
12
More intriguing information
1. The name is absent2. Financial Development and Sectoral Output Growth in 19th Century Germany
3. Empirically Analyzing the Impacts of U.S. Export Credit Programs on U.S. Agricultural Export Competitiveness
4. Draft of paper published in:
5. Constrained School Choice
6. Can genetic algorithms explain experimental anomalies? An application to common property resources
7. Job quality and labour market performance
8. The name is absent
9. The name is absent
10. Nonparametric cointegration analysis