Data analysis
Surveys that were incomplete or had the exact same
response throughout were removed (37 surveys, 10.66%).
Surveys marked non-random were also removed (26 surveys,
8.39%). Any scores for negative statements presented in the
survey were inversed to match the responses of the survey’s
positive statements. For example, high agreement with a
score of seven on a negative statement, “I am concerned
about elephants in zoos” was converted to a score of one to
keep coding consistent with that of the positive statements.
Similarly, a six was recoded as a two, or a three as a five.
Perceived elephant behavior data were transformed to
examine perceived activity levels and behavioral diversity.
The numerical values for active behaviors were calculated by
taking the number of active behaviors the guests reported
seeing, and dividing that by the number of total active
behaviors from which to choose from. The numerical values
for behavioral diversity were calculated by taking the
number of different species-appropriate behaviors reported
by zoo guests and dividing by the number of total speciesappropriate
behaviors possible. Principle component analysis
(PCA) with a Varimax Rotation, was chosen to assess
underlying patterns within the observed variability of the
major components, or factors, of the data set [Wielebnowski,
1999]. PCA allows for linear transformation of variables in a
smaller set of uncorrelated variables that remains representative
of the original data set [Dunteman, 1989]. Those
components with eigenvalues >1 were selected for
interpretation. The two highest positive loading scores
were used to label each component. Spearman Rho was used
to examine relationships among variables. SPSS (version
22.0 [Chicago, IL]) was used for all statistical analysis and
statistical significance was set at P