[1] The purpose of this reserch is to make clear what type of design can give good effects to win the high readership score. At the begining I had to know what factors decide the readership score, and how to change the quality of the desing into numerical value. I have used the method of multi-dimentional analysis with the help of a computer, and got the estimation formula, with 7 kinds of factors, as follows ; A^^〜=
B
+
C
+D+K+aE+bF+cG. A^^〜=estimated values of the readership score. 1)
B
,
C
, D, K : these four factors are given as category values (so we call them); each factor is classified into several categories.
B
=week day, classified into 7 categories.
C
=pages of a newspaper, which are characterized by the kinds of articles, news, and the other reading matters, classified into
22
categories. D=kinds of goods and trades of the advertizement, classified into 14 catedories. K=constants, given to each 13 surveys. 2)
E
,
F
, G : these factors are given as numerical values. G=area of advertizement, measured by the unit of column in the whole page length, being classified into
6
kinds of area (2.
5
, 3.
5
,
5
, 7, 10, 15) and 15 column means a whole page, about 2,000cm^2.
E
,
F
are the marks obtained in the design of advertizements.
E
=sum of the marks obtained in each design element on the advertizement.
F
=the marks obtained on the whole effect of the design. 3) a,
b
,
c
: coefficients Using the above formula with the category values and the coefficients, we can obtain the naked design effect from the actual value A of readership score: aE+bF=A-(
B
+
C
+D+K+cG). In this research, I used 1,829 data of readership score from 1960 to 1968, being obtained by 13 survevs, spring and autumn twice a year. The sample of size each survey was 3,541, 2,251, or 200 in the other 11 surveys. Multiple correlation coefficient between the estimated value A^^〜 and the actual value A is
0
.951. The table 2^* shows the contribution indexes of each factor by 3 kinds of expression-(1) Range : the absolute difference between the maximum and the minimum category values, (2) Standard deviations of the category values, (3) Partial correlation coefficient : the relationship between the actual values A and each factors. *see the table 2 in the thesis in Japanese. [2] The important point was in the determination of the values of
E
,
F
.
E
is the sum of the marks obtained in each design element, classified into four kinds :
E
=
E
_1+
E
_2+
E
_3+
E
_
4
. design elements [table]
E
_1,
E
_2,
E
_3,
E
_
4
, these values have the
5
grades, as
0
, 1, 2, 3,
4
and
0
is given to the design that has no attractive effect or no applicable element. And
E
obtain the marks of
0
to
9
grades as the sum of them. The values of
F
have also
5
grades,
0
to
4
. The principles to determine the values of
E
,
F
are as follows : a) The marks obtained of
E
,
F
must be the relative values among the each survey, and at the same time, they must have constancy within the same survey-the same elements of design must win the same marks obtained.
b
) They must be determined as to win the highest multiple correlation coefficient, when they are put into the estimation formula.
c
) They must be reasonable. In order to justify them, we must carry many researches on the actual condition. The frequency of accurence of each grade of
E
,
F
values, as the table
5
-
4
, and
5
-
5
in the thesis in Japanese.
抄録全体を表示