eldorado.tu-dortmund.de/server/api/core/bitstreams/a427fefe-bbe5-4ead-aea2-533d2ea7844e/content
given by
λ1 = a− b+ c− d
2 +
√ (e− f)2 +
( c− d− a+ b
2
)2
,
λ2 = a− b+ c− d
2 −
√ (e− f)2 +
( c− d− a+ b
2
)2
,
λ3 = a+ b+ c+ d
2 − e− f.
Using these results, it is possible to identify with a numerical search [...] that,
we use Proposition 2 of Kunert (1983) which claims that
C̃ (1) d = B2M
T d ω
⊥([P,U,Td,Sd])MdB2 ≤ B2M T d ω
⊥([U,Td,Sd])MdB2 = C (1) d , say,
with equality if and only if
(MdB2) Tω⊥([U,Td,Sd])P = 0. [...] multiplication with a matrix Bq = ω⊥(1q) without
changing the result:
C̃ (1) d = B2C̃
(1) d B2,
C̃ (2) d = B4C̃
(2) d B4.
5.2.1 Mielke and Kunert (2018)
Since the row and column sums of C̃ (1) d are 0, the …
