2.3. Statistics
For statistical calculations frequency distribution tables and
multiple responses tables were used. Cluster analysis was made
using Kohonen's artificial neural networks. This allowed to classify
surveyed businesses in clusters of similar opinion on HACCP system.
The purpose of cluster analysis was to arrange objects into groups of
similar characteristics (Kohonen, 2001). When m variables are
measured for each of n objects, these can be represented by n points
in m-dimensional space. This allowed to describe the similarity or
dissimilarity of these objects when the distance between the objects
in multi-dimensional space was calculated. The distances between
samples were defined by Euclidean distances computed from the
data obtained in the surveys. Analysis of variance was used to
determine the similarities of two clusters to be linked together. Data
were analyzed using Statistica ver. 10 software.