2.4. Statistical Analysis

Descriptive statistics were used to summarize the sample characteristics. Considering that both ADKS and DAS contribute to dementia literacy, we put two continuous variables into the multivariate analysis of variance (MANOVA) model as the dependent variables. Two-way MANOVA was used because two or more independent variables could be tested simultaneously with two dependent variables. Demographics (gender, education level, employment level, and accommodation) and location (Hong Kong, Macau, Zhuhai, and Guangzhou) were the independent variables. Interaction terms between each demographic factor and location were also put into the model to assess the effect of demographics on dementia literacy, in different cities.

MANOVA was used after validating the normality and homogeneity assumptions. The existence of multivariate outliers was assessed by a test of normality, and outliers were eliminated, when detected. The normality of the data was also assessed using the Shapiro–Wilk test; normality was assumed at p > 0.05. We then assessed the Mahalanobis distance of the two dependent variables (ADKS and DAS). Since the Mahalanobis distance of this data was 13.44 (that is, less than the critical value of the Chi-square of 13.82), multivariate normality was assumed. According to Levene’s test (p > 0.05), the criterion group variances were homogenous. We assessed the maximum likelihood criteria (Wilk’s lambda), and effect size was presented as partial eta (η²). When the MANOVA detected significant statistical differences between the IVs and the DVs, we proceeded to conduct a two-way ANOVA for the dependent variables and a post-hoc test, to assess the effect of categorical independent variables on the dependent variable.

Four logistic regression models, using a Wald stepwise backward elimination method with the probability of a stepwise entry alpha of 0.05 and removal alpha of 0.10, were conducted to identify the associated factors (demographics, the cities where the participants lived, dementia knowledge, and attitudes) in favor of a particular type of mass media, for receiving information on dementia. All statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS 25.0, IBM Corp., Armonk, NY, USA). The level of significance was set at p < 0.05.