Comparison in oral and manual FA and oral, manual and oral-manual FP across extent of encountering packaged food under experimental conditions were done using the test of multiple proportion. For similar comparisons in the non-experimental scenario, we either applied multilevel binary logistic regression or permutation test depending on fulfillment of statistical criteria. Test of multiple proportions was carried out in Statistics Toolkit (StatsToDo, 2020), binomial logistic regression and permutation test were done in RStudio v.1.4.1717(RStudio Team, 2021) and effect size index (Cohen’s d ) was computed in an online effect size calculator (Wilson, 2001) based on Lipsey and Wilson (2000).
While analyzing data from naturalistic settings (non-experimental condition) for FA and FP, we treated each species separately. Due to near complete separation of hand and mouth use in FA along encounter with packaged food and urbanization in all species except vervet monkey, multilevel regression generated singularities and therefore, a permutation-based chi-square test was applied with chiperm()function in the GmAMisc package (Alberti, 2021). Hypotheses testing in vervet monkeys (1) modeled the distribution of FA as negative binomial using glmmTMB() function in the glmmTMB package (Brooks et al., 2017), (2) performed model diagnostics usingtestDispersion() and simulationOutput() functions in theDHARMa (Hartig, 2021) package and (3) selected optimal model based on Burnham and Anderson (2002).