*2.4. Analysis*

We conducted a hierarchical cluster analysis to partition the sample into mutually exclusive groups based on broad patterns of in-home drinking water consumption. We used Ward's minimum variance method, which groups items into clusters based on the similarities and differences between each data point [56]. To determine an optimum cluster solution to use in subsequent analysis, we used both the Calinski/Harabasz test and the Duda/Hart test to assist in identifying an optimal cluster solution [57].

We explored the internal consistency and dimensionality of multiple indicators using Cronbach's alpha and factor analysis. We conducted one-way analyses of variance (ANOVAs) to look for differences across in-home drinking water patterns based on trust, salience, risk, and water quality evaluation perceptions, in addition to a resident's drinking water behavior. We report comparisons of cluster mean differences (ANOVA contrast) and standardized mean differences (Cohen's d is a quantitative measure of the magnitude of the effect). A general guideline for interpreting effect size for Cohen's d is that a standardized difference <0.20 represents a small effect, a difference of 0.50 represents a medium effect, and a difference >0.80 represents a large effect.
