Professional Documents
Culture Documents
MARCH,1949
OF GROUP STRUCTURE
1. Introduction
I n a n u m b e r o f b r a n c h e s of t h e social sciences one e n c o u n t e r s
p r o b l e m s of t h e a n a l y s i s o f r e l a t i o n s h i p s b e t w e e n t h e e l e m e n t s o f a
grouP. F r e q u e n t l y t h e r e s u l t s of t h e s e i n v e s t i g a t i o n s m a y b e p r e sented in d i a g r a m m a t i c f o r m as s o c i o g r a m s , o r g a n i z a t i o n c h a r t s ,
flow c h a r t s , a n d t h e like. W h e n t h e d a t a to b e a n a l y z e d a r e s u c h t h a t
a d i a g T a m of t h i s t y p e m a y be d r a w n , t h e a n a l y s i s a n d p r e s e n t a t i o n
of t h e r e s u l t s m a y b e g r e a t l y e x p e d i t e d b y u s i n g m a t r i x a l g e b r a , T h i s
p a p e r p r e s e n t s s o m e of t h e r e s u l t s o f a n i n v e s t i g a t i o n of t k i s a p p l i cation o f m a t r i c e s . I n i t i a l t r i a l s in t h e d e t e r m i n a t i o n of g r o u p s t r u c t u r e s i n d i c a t e t h a t t h e m a t r i x m e t h o d is n o t o n l y f a s t e r b u t a l s o less
p r o n e to e r r o r t h a n m a n u a l i n v e s t i g a t i o n . *
T h e second s e c t i o n of t h i s p a p e r p r e s e n t s c e r t a i n c o n c e p t s u s e d
in t h e a n a l y s i s a n d a s s o c i a t e s m a t r i c e s w i t h t h e g r o u p in q u e s t i o m
The third states the results obtained and the fourth gives illustrations o f t h e i r a p p l i c a t i o n . F i n a l l y , s e c t i o n five c o n t a i n s a m a t h e m a t i c a l f o r m u l a t i o n o f t h e t h e o r y a n d d e r i v a t i o n of t h e r e s u l t s p r e s e n t e d
in section t h r e e .
2. Definitions
2.01. T h e t y p e s o f r e l a t i o n s h i p s w h i c h t h i s m e t h o d will h a n d l e
a r e : m a n a chooses m a n b as a f r i e n d , m a n a c o m m a n d s m a n b, a s e n d s
m e s s a g e s to b, a n d so f o r t h . Since in a g i v e n p r o b l e m w e c o n c e r n
*Some of these examples have been worked out by the Research Center for
Group Dynamics, Massachusetts Institute of Technology, in conjunction with
some of its research.
95
96
PSYCHOMETRIKA
ourselves with one sort of relation, no confusion arises f r o m ~eplacing the description of t h e relationship by a symbol : > . Thus, instead of " m a n i chooses m a n ] as a f r i e n d , " we w r i t e " i : > ]." If, on
t h e other hand, m a n i had not chosen m a n ], we would have w r i t t e n
" i if: > ] , " using t h e symbol :~" > to indicate the negation o f the relationship denoted by ~--- > .
2.02. Situations such as m u t u a l choice of f r i e n d s or two-way communication would thus be indicated by i ---~ > ] and ~
> i, or briefly,
i < - - > ]. We describe such situations by s a y i n g t h a t a s y m m e t r y
exists between i and ].
2,03. When the choice is not mutual, t h a t is i -~ > 3" or ] ~ > i
but not both, we say an a n t i m e t r y exists between i and ].
2.04.* The data to be analyzed are presented in a m a t r i x X as
follows: the i,] e n t r y (x~) has the value of 1 if i ~ > ] and the value
0 if i > ]. F o r convenience we place the re&in diagonal terms equal
to zero, i.e., x~ ~--- 0 for all i. This convention, i > i, does not restrict the applicability of the method, since there is little significance
in such statements as " J o n e s chooses himself as a f r i e n d . "
Suppose, f o r example, t h a t we had a gToup of f o u r members w i t h
the following relationships: a ~--- > b, b = > , b ~--~> d, d ~ > b,
c ~ > a, c ~ > b, d ~ > a, and d ~-~ > c. All other possible combinations of a,b,c, and d are related by the symbol ~ > . The X m a t r i x
associated with this group is:
b
c
d
1
1
1
0
1
1
0
0
1
1 ~
0
0
2.05. F r o m the X m a t r i x we e x t r a c t a symmetric m a t r i x S having entries s~j determined by s~- = ss~ - - 1 if x~i ~---xj~ -~ 1, and otherwise s~j ~ sj~ ~-~ 0 . All the s~mmetrie---s in t h e group are indicated in
the m a t r l x S . The S m a t r i x associated with t h e above X m a t r i x is:
0
1
0
0
1
0
0
1
0
0
0
0
0 7
1
0 |
0 J
*In the course of the present work it was brought to our attention that in
97
i--- > a , a ~
> b ,...,p~
2.07. When two n-chains have the same elements in the same
order, i.e., the same members, then t h e y are said to be equal, and
otherwise t h e y are distinct. I t is impol'tant in this definition of equality t h a t it be recognized t h a t both .the elements of the group and t h e i r
order in the sequence are considered. The two chains i,],k,l,p and
i,p,k,],l are distinct though t h e y contain the same five elements.
2.08. When the same element occurs more t h a n once in an
n-chain, the n-chain is said to be redundant. (Thus, in a group of
elements any n-chain w i t h n g r e a t e r t h a n m is r e d u n d a n t ) . The chains
a,b,e,d,b,c and a,c,a,b,d,c,e are, for example, both redundant, for the
element b occurs t~dce in the f o r m e r and the elements a a n d c both
occur twice in the latter. An example of a non-redundant 5-chain is
a,d,p,b ,q,] .
2.09. A subset of .the group f o r m s a c,lique provided t h a i it consists of three or more members each in the symmetric relation to
each other member of the subset, and provided f u r t h e r t h a t there
can be found no element outside the subset t h a t is in the s y m m e t r i c
relation to each of t h e elements o f the subset. The application of this
definition to the concept of friendship is immediate: it states t h a t a
98
PSYCHOMETRIKA
set of m o r e than two people form a clique if they are all mutual
friends of one another. In addition, the definition specifies that subsets of cliques are not cliques, so that in a clique of five friends we
shall not say that any three form a clique. Although the word "clique"
immediately suggests friendship, the definition is useful in the study
of other relationships.
2.10. This definition of clique has two possible weaknesses:
first, if each element of the group is related by ~- > to no more than
c other elements of the group, then we can detect only cliques with
at most c + 1 members; and second, there may exist ~Sthin the group
certain tightly knit subgroups which by the omission of a few s>nnmetrics fail to satisfy the definition of a clique but which nonetheless
would be termed, non-technically, "cliques." It may be possible to
alleviate these difficulties by the introduction of so called "n-cliques"
which comprise the set of n elements which form two distinct n-chains
from each element of the set to itself. This requires that the n-chains
be redundant with the only recurring element being the end-point
and also that all the relations in the n-chains be symmetr/c.
This definition means that the four elements a,b,c, and d form
a 4-clique if the 4-chains (for example) a,b,c,d,a and a,d,c,b,a, both
exist. These by the definition of n-chain require that the relations
a<=>
b, b < - - > c ,
c<-->
d, d < - - > a
exist, but nothing is said about the relations between a and c, and
b and d . The original definition requires, in addition, that
a<=>e
and
b<=>d
for a,b,e, and d to form a clique of four members. Thus we see that
the definition of n-clique considers "circles" of symmetries, but it fails
to consider the symmetric "cross" terms that exist between the members of the n-clique. These cross terms will be investigated, however,
by determining whether any m of these n-elements form an m-clique.
The usefulness of the definition of n-clique can be judged only
a f t e r experience has been gained in its application. This is not conveniently possible at present, unfortunately, because the problem of
the general determination of redundant n-chains has not been solved
(see 5.09).
The most general definition of a elique-like structure including
antimetries will not be discussed, for it is believed that this will not
be amenable to a concise mathematical formulation.
3. Statement of Results
3.01. In X ~ the entry x~j(~) ~ c if and only if there are c dis-
99
100
PSYCHOMETRIKA
s,~9(~) =
+
+
--
[ (4 - - 2) (4 - - 1) - - (0 - - 2) (0 - - 1)]
[(3--2) (3--1)-- (1--2) (1--1)]
[(5--2)(5--1)--(3--2)(3--1)]
+2
18 .
X=
X~__=
O
0 10
1 0
01
0
11
0
00 1
1
1
1
0
0
0
0
1
0
I
I
I
0
1
2I
1
1
0
00
2
2
1
01
0
0
1
01
1
1
2
21 1
0
,
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
To obtain the m a t r i x of r e d u n d a n t 3-chains we compute the following matrix, in which the symbol R(:~) stands for the matrix of red u n d a n t 2-chains:
X R c:2~ + R c 2 ~ X ~ S .
Deleting in this sum the main diagonal and replacing it by the main
diagonal of X 3 gives the m a t r i x of r e d u n d a n t 3-chains (5.09). If
t h e main diagonal of X R ~ + R ( = ~ X - - S is denoted b y y(3) and the
main diagonal of X C3) by Z (3), then let E ~3) = Z TM - - Y(a) and thus
the m a t r i x of r e d u n d a n t 3-chains, R (3), is given b y
R ~3~ =
X R ' ~ + R c 2 ~ X + E TM - - S .
101
4. Examples
4.01. As the first example, let us compare and analyze the friendship structure in the t w o following hypothetical groups. The matrices
are (where a blank entry indicates a zero) :
g,A9 ~"
l,--i
b~
....
p~
oo
,.,.i
c~
d~
d)
b~
b~
~b
~s
~.~
--
I-~
-',-1
e~
-q
t~
t~
~b
~.
r
~
~--~
~,-L~
--
~w
i~ .
i-.Ll-~l.~
i-~
~,~.
~..~l..~i,~
;J
O0
,..I
b~
O0
-.I
b~
~J
b~
103
1
3
4
5
6
7
8
9
10
I
1 2 3 4 56
1415
14
1514
14
7 8 9
1412
1412
10
12
12
1414
13
10
10
1
1414
1212
1212
13
8
10
lr
2
3
,
1010
10 6
8
5
6
II
12345678910
2
564
2
241
5
421
11
4
8 j
7
6
11
11 1
41
1
I
I
2
9
10
1
4
1
141
11
2
2
1
104
PSYCHOMETRIKA
overlaps the previous two. All the men contained in cliques have been
accounted f o r at least once, and a check either with the formula f o r
main diagonal entries (3.05) or directly in the S m a t r i x indicates
that all the cliques have been discovered. This, coupled w i t h w h a t
we discovered in S ~, completely determines the symmetric s t r u c t u r e
of the first group.
F o r purposes of qualitative judgznent and a guide to carrying
out analysis, we note that the first t w o rows of S 3 present an interesting s u m m a r y of the clique structure. The entries sl~ (3) and s2~(3)
have the largest values, next largest are in columns f o u r and seven,
and then finally in columns eight and ten. Men 1 and 2 a r e contained
in all three cliques, 4 and 7 are each contained in t w o cliques, and
finally men 8 and 10 are each in only one clique. This indicates that
the magnitude of the off-diagonal t e r m s determines to some extent
the amount and structural position of the overlap of cliques.
h group II there are only three elements with non-zero maindiagonal entries, all with the value 2. This fits the formula (t--2)
( t - - l ) with t ------ 3 (3.04). Thus the men 1, 3, and 4 form a clique
of three members. R e t u r n i n g to S ~, we see t h a t t h e r e remains one
unaccounted s y m m e t r y each for men 4 and 9, hence 4 < - - > 9 .
In group I, the off-diagonal terms a r e large in magn.itude and
are quite dense in the array, w i t h some rows completely e m p t y or
with single entries in the S 3 matrix. This indicates a closely knit
g r o u p wit,h certain men definitely excluded. The S ~ m a t r i x f o r the
second group has f e w e r entries of a smaller value indicating a tess
tightly knit structure, b u t it has no e m p t y rows and only one row
with a single e n t r y ; t h a t is, it has f e w e r people t h a n gToup I who
are not accepted b y t h e group or w h o do not accept it.
A consideration of the mat~;ix X ~ S will give all t h e antimetries
in the groups and complete the analysis of the structures.
It. is clear t h a t this procedure gains s t r e n g t h as the complexity
of the problem increases, for the analysis of a twenty-element group
is little more difficult than t h a t of a ten-element group.
4.02. The second example is a communication system comprising t w o - w a y links between seven stations such as might occur in a
telephone or telegraph circuit. The n u m b e r of channels of a given
number of steps (i.e., n-chains in the general theory) between any
t w o points and the minimum n u m b e r of steps required to complete
contact between t w o stations will be determined. Suppose the matrix
of one-step contacts is:
1234567
11
2
3
4
5
6
7
1-
11
11
1
111
11
11
11
1
111
105
-3
112210
211101
141123
114322]
113422j
1 022232
0 132224
1
1
2
2
(:2)
F r o m the former, the two connections 1 < ~- > 7 and 2 < ~ > 6 cannot be realized because x17~) ~ x712) ~- 0 and x2+(~> ~ x+~<2) - - 0
(3.01). The contacts a r e possible i n - ~ r e e steps, however, since~ X 3
is completely filled. Thus two steps a r e sufficien~ for most contacts
and three steps f o r all.
In dete~Tnining the n u m b e r of paths between t w o points it is desirable to eliminate r e d u n d a n t paths: F o r two-step communication
this is done by dele*ing the m a i n dia~onal of X ~. The r e m a i n i n g t e r m s
represent the n u m b e r of two-step paths between the stations indicated. The m a t r i x of redundancies f o r three-step communication is
given by R ~+~ "-- X R ~2) ~+ R ( ~ > X + E ~> - - S (3.06), which works out
to be:
106
PSYCHOMETRIKA
1
2
3
4
5
6
7
1234567
246
6
425
65477
78767
77867
6656
6
7766
R(~):
1234567
1
2
3
4
5
6
2 43434323J
2
3355
433
234
4332
34
43533
2J
235442
We notice that the three-step paths between I and 2 and 2 and 3 are
all redundant but that there are two-step paths for these combin~tions. All other combinations have at least two t~ree-step paths joining them.
5. Mathematical Theory
5.01 To c a r r y out the following mathematicM formulation and
the proofs of theorems it is convenient to use some of the symbolism
and nomenclature of point set theory. As there is some diversity in
the literature, the symbols used are:
Sets are either defined by enumeration or by properties of the
elements of the set in the form: symbol for the set [symbols used for
elements of the set I defining properties of these elements]. When
a single element i is treated as a set it will be denoted by (i), otherwise sets will be denoted by upper case Greek letters.
The intersection of (elements common to) two sets P and
is denoted by F .
The union of two sets F and (elements contained in either or
both) is denoted by P + ~ . The context will make it clear whether
the symbol i+ refers to addition, m a t r i x addition, or union.
The inclusion of a set P in another set (all elements of P are
elements of ) is denoted by P < . The negation is P <* .
If < 2 , then the complement of with respect to 2 , % is de,
107
if
if
i=j
i=#]
2. i ~ = > i .
Let a n u m b e r x~j be associated w i t h i and 3"such t h a t
--1
----0
if
if
i-->]
i>].
if x ~ j = x j i = l ,
otherwise.
i.e.,
i<:>].
108
PSYCHOMETRIKA
2. Two n - c h a i n s F and a r e e q u a l if and only if the r ta member of F equals the r th member of ~ , i.e., 7~ - - ~, f o r 1 < r < n + 1.
I f this is not true, t h e n / " and are d i s t i n c t .
3. Each pair of elements 7~ and 7~ of a n n-chain with 1 < k <
m < n + 1 a n d 7~ : 7.~ is said to be the r e d u n d a n t p a i r ( k , m ) . A n
n-chain is r e d u n d a n t if and only i f it contains a t least one r e d u n d a n t
pair.
4. The elements 1 , 2 , - . . , t (t > 3) f o r m a c l i q u e 0 o f t members i f a n d only if each element of O is s y m m e t r i c w i t h each other elem e n t of 0 , and .there is no element not in O s y m m e t r i c w i t h all elements of O .
This is equivalent .to
x i j = 1 - - &j f o r / , ] = 1 , 2 , . - . , t but not f o r i , ] =
t + 1 , whatever the (t + 1) ~ element.
1,2,.-.,
t,
X p q - - - Xqi =
1 ,
and if the indices were not so selected then a t least one x~, ~-- 0 . Thus
n-chains contribute 1 to the sum and other ordered sequences cont r i b u t e 0 . Since the indices take on each possible combination of
values j u s t once; every distinct n-chain is represented j u s t once. If
there are c such n-chains, then t h e r e a r e a total of c ones in the summation.
5.05. Theorem 2: An element of ~ has a m a i n diagonal value
of c in X ~ i f a n d only i f it is s y m m e t r i c ~'ith c elements of 2 .
P r o o f : Let ~ be the set of ]'s for which i < ~- > ]. B y definition
x . (~) =
E x~jxj~ + E x ~ x j ~ = ~
je~
~+ ~,~.
je"
~]~ " - c by theorem 1 (5.04) and N~ - - 0 because i and ] are not symmetric f o r ] s ~', so either x~j = 0 or xj~ ~---0 or both. Thus i f i is symmetric with c elements of 2 , x~j (~) : c .
109
S e l e c t ] and k s u c h t h a t
(]) :+ (k) < O and such t h a t i # ] #
k
# i . Such elements exist b y t h e definition of a clique (definition 4,
5.03). I t is t r u e b y t h e definition o f a clique a n d of the m a t r i x S
that: s~s ---- sj~ = s~ - - s~j : s~ = s~ - - 1 f o r such ] and k . Thus this
choice of ] and k contributes 2 to--~e summation, and because s~j > 0
for all i and ] t h e r e are no negative contributions to the s u m ; therefore s~ (~) > 2 > 0 .
Suppose t h a t s~ (8) > 0 . Then there exists at least one p a i r of
elements of ] and k such t h a t s~i = sj~ ---- sk~ ~--- 1 a n d this implies
i <=>
3",3' < - - >
k , and k < = >
i . I f there a r e no other elements symmetric w i t h i , ] , and k then these t h r e e f o r m a clique. I f
there is another element s y m m e t r i c with these three, then consider
the set of f o u r f o r m e d b y adding it to the previous three. I f t h e r e ~s
no other element s y m m e t r i c with these four, they f o r m a clique. I f
there is, add it to the set and continue t h e process. Since the set 2
contains only a finite n u m b e r of elements, the process must terminate
giving a clique containing i .
5.07.
zl~ ~--- O~ (e~ :+ 0~ + -.. + O~_~) have d~ members, and 3) i is contained in the cliques @~, z - - 1 , 2 ,
--., m , then
Proof: By definition
- -
(i)+(k)<Y~
The set of all the pairs ] , k is the union of the following three mutually exclusive sets:
T~ [ ] , } I t h e r e exists v such t h a t (]) + (k) < 0 , ; there does not
exist ,a such t h a t (]) :+ (k) < 0 ~ , (]) + (k) <" A~]
110
PSYCHOMETRIKA
F o r T1 then either
a)
(j)
ore)
(j) + (k) < O~ if and only if (j) + (k) < A ~ f o r a l l a .
This is not possible because A~ = 0 . Thus gJ~ is empty.
2. (j) + (k) < T2 impties'suss.~s~ = 0 for SuSjkSkl ~-- 1 implies t h a t i ; ] , and k are either a clique or a subset of a clique (by
t h e a r g u m e n t of theorem 3), but (j) + (k) < gz implies ] and k
are not contained in a n y clique.
3.
T~ gives t h a t
(j)+ (k) <q,~
= Z
V=l
s~jsi~s~
(j)+(k)<@ v
(j)+ (k)<*A v
I (])+
(k) <
o.~, (y) + (k) < A,] + Q~[S,k i (]) + (k) < O,, (y) + (k) <* Av]
and since .% - 2~ = 0, it follows t h a t Y~ = Y~ + N or Y~ = N -- ~ .
~h
Ft2
fl~
~3
f~l
fte
t?, is the set of all ordered pairs (j) + (k) < O,,. I f i C j C k i ,
then s~ = s~k = s~ = 1 , othetavise one of the s~ - - 0 . Since every
O, contains t,, elements, there are %-aP~ ordered pairs s a t i s f y i n g these
conditions. Thus:
Similarly
z={o
' (d'o 2).
I)
11since
AI--O.
Combining these,
~a
(dv--2)(d~--l),~>l
, v -- 1.
Iii
S u m m i n g o v e r ~, gives
s.~ (~) = ~ ( (t, - - 2) (t~ - - 1) - - (d~ - - 2) ( & , - 1) }
Y=2
+ (tl - - 2) (tl - - 1)
Since the entries szj(3) are uniquely determined from the entries
of S by the laws of matrix multiplication, all valid methods of calculating s~ (a) will give the same result. Specifically, in the above
formula the numbering of the cliques is immaterial.
Similar formulas to that just deduced may be given for the offdiagonal terms of S 3, but they are considerably more complex, and,
to date, they have not been found useful in applications.
5.08. T h e o r e m 5: I f 1) 0 is a set of t m e m b e r s w i t h t > 3 ,
2) s~/~) ~ ( t - - 2 ) ( t - - l ) f o r i c o n t a i n e d in O , a n d 3) sj/~) = 0 f o r ]
c o n t a i n e d in O', t h e n O is a clique of t m e m b e r s .
P r o o f : T h e r e a r e t w o cases:
1. i < = > 2" f o r all i , 2" s O , t h e n 0 is a clique b y definition 4
(5.03) and t h e o r e m 3 (5.06), and it ,has t m e m b e r s b y p a r t 1 of
the hypothesis.
2. T h e r e e x i s t p a n d q s O such t h a t p a n d q a r e n o t s y m m e t r i c .
T h e n b y definition
S~/~) =
Z Z
s~jsjks~
U)+(~)<o
Z E
s~jsj~sk~.
(j)+(k)<*O
S u (~) - -
Y. Y~ si~sj~s~i - - 2 = ( t - - 2 ) ( t - - l ) - - 2
(j)+(k)<~
112
PSYCH0~ETRIKA
Redundancies:
(i,]) =
o.
2. The B, class: There exists one and only one redundant pair
(k,m) and it has the property:
(k,m)
3.
(i,]) = i .
The C~ class: There exists one and only one redundant pair
4. The D,~ class: There exist two and only two redundant pairs
(k,m) and (p,q) and they have the properties:
(k,m)
(i,D = i
( p , q ) (i,]) = ] .
5. The E, class: There exists one and only one redundant pair
(k,m) and it has the property:
(k,m)
(i,j) =
(i,i).
113
R. D U N C A N L U C E A N D A L B E R T D. P E R R Y
R ( z ) r-- 0
A (8) - - 0 .
(p,q)
(i,D
= i
(i,]) -" ].
t h a t is,
--1
d~/8) : 0
if i < = > ]
otherwise,
~ = yl , y~ , ~ ,
(b)
y~ , - - ,
(c)
y,~ , - - ,
7~ , ~'~,l-=
wherea+b+c+5=n,k<m,andyk=y,~.
(~t-2)
72
> ~ ----- q is a r e d u n d a n t
chain d e t e r m i n e s no m o r e t h a n
to ] . T h u s t h e n u m b e r of re] is t h e sum o v e r all combinaof r e d u n d a n t n : - 2 chains f r o m
(p)+(q)<,~
or
A (~) = X R ( ~ ' 2 ) X .
114
PSYCHOMETRIKA
I f the m a t r i x [e~/*)] is defined as
[e,/-)] = X R ( , ~ ) X
+ D oo
TM)
+ R(~-I)X _ XRO~2)X
+ EOO - - D(~) .
Uniqueness:
115
Since X is non-singular, IX~I 0 , and t h u s there exists a unique inverse of X % X -~q, such t h a t X ~ X bq = I ; therefore X = Y .
5.12. Theorem 7: I f n is a positive odd integer and S a real
symmetric matnix, then there is one and only one real symmetric n ~I'
root of S .
Proof: 1. There is one such n t~ root.
Since S is real and s~m~netric there exists a real orthogonat
m a t r i x P such t h a t P ' S P = D ( P ' is the transpose of P) is diagonal
with real entries d~ which are the characteristic roots of S .* Assume
P is so chosen t h a t dl~ ~< d~ < .-. < d ~ . L e t B be the diagonal matrix of the real n tj' roots of the elements of D , i.e., b~ - - real ( d ~ ) ~/" ,
so
B~=D.
Define R = P B P ' .
(1)
Then R " = S , f o r
R ~=
n - -
- - Q R !~ Q ~ - Q ' S Q
= (P'Q)'D(P'Q)
(2)
Substituting (3) in
(3)
(2)
D--
U'DU
or
UD "- DU.
116
PSYCHOMETR!KA
or
u~k (d,k - - d . )
= 0.
(4)
Since the dkj: are real and n is odd, equation (4) implies
u ~ [ ( G ~ ) v~ - - ( d . ) ""] = 0
(5)
B = u'Bu.
,~ ,
Q'PBP'Q
or
QTQ' ~- PBP'.
But
Q T Q ' ---- R1
and
PBP'
= R
by definition ; therefore
RI ~ R .
6. A c k n o w l e d g e m e n t