AD ALTA
JOURNAL OF INTERDISCIPLINARY RESEARCH
Size of company is defined according to regulation of European
Commission (Table 1). According to selected NACE groups, the
basic population has been defined for individual country as
follow in Table 2. Sample population of agriculture companies
consist 4996 from whole Europe (see Table 2).
Table 1 Limits for splitting of companies into individual
categories
Staff
headcount
Annual
turnover
Annual sum
of balance
Micro
< 10
≤ 2 mio €
≤ 2 mio €
Small
< 50
≤ 10 mio €
≤ 10 mio €
Medium
< 250
≤ 50 mio €
≤ 43 mio €
Source: Evropské společenství, 2006
Table 2 Pivot table: company size and European region
Micro
&
Small
Medium
Large
Missing
Total
Western
105
24
7
159
295
Southern
2259
304
41
120
2724
Northern
134
20
10
36
200
Eastern
1026
476
108
167
1777
Total
3524
824
166
482
4996
Source: own work by authors
Questionnaire survey as part of customers’ analysis was targeted
on field of outdoor clothes, knowledge the producers’ brands.
This survey was realised during spring of 2017 in Czech
Republic. From group of customers there were selected 851
respondents in random way to participate. From that amount 292
questionnaires were returned back (relative amount is 34,31 %).
Factor analysis is based on the selection of correlation and
partial correlation coefficients. The correlation coefficient
represents the closeness of linear dependence of individual
variables and partial correlation coefficients. The partial
correlation coefficient shows a similarity of two variables in
such a situation that the other variables are assumed constant. If
it is possible to explain the dependence of variables using
common factors, the partial correlation coefficients are very
small, close to zero. To assess the suitability of the factor
analysis, two tests can be used (Tarnanidis et al., 2015; Conti et
al., 2014):
Kaiser-Meier-Olkin (KMO) is a coefficient which could
reach values between 0 and 1. Its value consists of the rate
of squares sum of the correlation coefficients and squares
sum of the correlation and partial coefficients.
The use of Bartlett’s sphericity test lies in testing the null
hypothesis stating that the correlation matrix of variables is
unit (on diagonal, there are only ones, others are zeros). If
the null hypothesis is rejected, the factor analysis may be
used for the defined variables.
For the purposes of verification of the factor analysis Cronbach’s
alpha indicator must be used. This indicator is understood as a
reliability coefficient, used as a kind of analogy with the
correlation coefficient. Normally, values oscillate in the interval
〈0;1〉. Zero, as the extreme value, describes the situation in
which individual variables are uncorrelated. On the other hand,
the value 1 describes correlated variables. When the value is
closer to 1, a higher degree of conformity is reported (Hrach,
Mihol
a, 2006; Cronbach, 1951; Řehák, Brom, 2016).
However, high Cronbach’s alpha does not imply that the
measure is dimensionless. If, in addition to measuring internal
consistency, you wish to provide evidence that the scale in
question is dimensionless, additional analyses can be performed.
Exploratory factor analysis is one of the method to check
dimensionality. Cronbach’s alpha is not a statistical test; it is a
coefficient of reliability (or consistency). The value could be
expressed as the function of number of test items and the average
inter-correlation among the items. Below, for conceptual
purposes, we show the formula for the standardized Cronbach’s
alpha:
where N equals to the number of items; c-bar is the average
inter-item covariance among the items; v-bar equals to the
average variance.
The values of Cronbach’s alpha could be from 0 to 1. If the
values are close to 0.5, it signifies a bad level of internal
consistency. Over 0.7 means that the value is acceptable and
values close to 1 are excellent. A “high” value of the alpha is
often used (along with substantive arguments and other
statistical measures) as evidence that the items measure an
underlying (or latent) construct (Hinton et al., 2004).
Correspondence analysis describes relation between both two
nominal variables in pivot table and individual categories. In
pivot table there is category combination which should become
significant or not. If any categories are similar or associated,
there are located in graph near themselves. Correspond analysis
itself is focused on association rate, usually by chi-square
measure. There are nominal variables as input into correspond
analysis, and kind of premise, that there is no ordering between
variables (McGarigal, Cushman, Stratford, 2000; Beh, 2010,
2008). Correspond analysis processes dimensional homogenous
data which consist only positive values or zeros. Chi-square
range has become coefficient which excludes zeros, and help to
define relations between rows and columns.
Calculation of correspondence analysis includes three steps: (1)
pivot table transformation into table with support of Pearson chi-
square; (2) individual value decompositions are applied into
defined table, then there are calculated new values and new
vectors; (3) new matrix operations serve as input to graph
design. Basis for two dimensional pivot tables is data matrix
n×2, in which categorical variable A get r values (a
1
, a
2
, .. a
r
)
and categorical variable B get s values (b
1
, b
2
, .. b
s
). Due
realised observation there is created table by two dimensional
separations of both variables. In the table is used n
ij
frequency,
which represents intersect of both variables. This n
ij
provides
number of observations, where are both a
i
and b
j
. Except n
ij
there are used marginal frequency n
i+
, where own observation
with a
i
value are observed (similar approach is for n
j+
in
column). After estimating the theoretical frequencies there is
designed chi-square statistics. This statistic has chi-square
distribution and number of degrees of freedom (r-1)(s-1). On this
basis, it is decided if exist dependency between variables in the
population, and by using correspondence analysis is also
possible to determine the structure of dependence (Beh, 2010;
Kudlats, Money, Hair, 2014).
4 Results
Based on the economic data from Amadeus database, it is
evident that companies commonly use traditional financial
indicators for measurement of their own performance. These
indicators were analysed:
x
1
– Cash flow [th EUR];
x
2
– P/L for period (Net income) [th EUR];
x
3
– Operating revenue (Turnover) [th EUR];
x
4
– ROA using P/L before tax [%];
x
5
– ROE before tax [%];
x
6
–Gross profit [th EUR];
x
7
– Shareholder funds [th EUR].
Based on the statistical characteristics of the examined groups
the conclusions could be presented as an approximate result,
limited by the resulting reliability. In the results of the paper
there are characteristics of research barriers and future research
possibilities.
- 145 -