Data analyses
We applied the Cronbach’s alpha coefficient to test the reliability and
internal consistency of responses across items in the questionnaire
(acceptable value= 0.80 > α > 0.70). We
characterized each visitor’s perception and attitude according to their
sociodemographic backgrounds (e.g., gender, group, age, educational
attainment). To test for the significant difference in categorical
responses based on their sociodemographic backgrounds, we used Pearson’s
chi-squared test of independence (ꭓ2). A Cramer’sV was then applied to measure strength of association between
variables (V values: weak < .10, low < .10> .30, and moderate < .30> .50).
We performed separate generalised linear regression (GLMs) to determine
variables that predict tourists’conservation willingness in the (i)
pre-visit and (ii) post-visit. We pooled the responses “Not sure” with
“No” and coded as “Not willing” and “Yes” as “Willing” (e.g.,
Carson et al. 1995; Mulema et al. 2020) to fit with the binary modelling
following Aziz et al. (2017). We used the values obtained from mean
scores of each indicator in pre-visit and post-visit variables
(Supplementary data 1) and sociodemographic profiles as covariates and
factors in the binomial logistic model. The (a) prior experience and
recognition, (b) knowledge and understanding about Philippine bats, and
(c) and prior knowledge of ecosystem services were used as predicting
variables for pre-visit conservation willingness. Next, we used: (a)
learning and aesthetic satisfaction, (b) perceptions about conservation
importance, (c) enhanced knowledge of ecosystem services of bats and (d)
prior knowledge about ecosystem services as predicting variables for
post-visit conservation willingness. We choose the best model based on
the lowest of Akaike’s Information Criteria (AIC) values.
Consequently, to determine the effectiveness of short-term enhancement
of knowledge about bat ecosystem services and conservation willingness,
we applied the non-parametric Wilcoxon rank test to assess the
significant relationship between visits. All reliability and statistical
tests were performed using the open-source program Jamovi version 1.2.6
(The Jamovi Project 2020). We set our statistical significance atp = .05.