2.4. Statistical Analysis

Data analysis was carried out with SPSS 26 (IBM, Athens, Greece). The five quality dimensions were calculated as mean values of the variables/questions that composed each one of them. This was done for both expectations and perceptions and thus ten new variables (five pairs of perceptions-expectations for each service quality dimension) were created. Afterwards, the gap between perceptions and expectations was calculated for these variables by subtracting expectations from perceptions [P-E]. Kolmogorov–Smirnov and Shapiro–Wilk tests were applied to assess normality of their distributions. These tests showed a statistically significant deviation from normality. Additionally, their graphical illustration using boxplots displayed many outliers for all the gaps, thus all the statistical tests that were used were non-parametric (Table 1).

Normality tests for SERVQUAL dimensions.

^a Lilliefors Significance Correction. GT: Gap Tangibles, GRel: Gap Reliability, GRes: Gap Responsiveness, GA: Gap Assurance, GE: Gap Empathy.

Specifically, the non-parametric Mann–Whitney U test was used to determine possible statistically significant differences of the gaps between two independent groups and Kruskal–Wallis H test was used to determine whether statistically significant differences existed between more than two groups (with post-hoc analysis based on the non-parametric Dunn’s test with Bonferroni correction). Additionally, Wilcoxon matched-pairs signed-ranks test was used to check for statistically significant differences between perceptions and expectations for each of the five pairs of the quality dimensions and the non-parametric Spearman’s Rho correlation coefficient was used in order to evaluate for possible correlations between the gaps of perceptions-expectations of the SERVQUAL dimensions.

Regarding the reliability of the questionnaire, Cronbach’s alpha coefficient was calculated separately for each section of the questionnaire which composed the quality subscales, for both expectations and perceptions. Its values ranged between 0.68 and 0.88, [Tangibles: (E) = 0.76, (P) = 0.80, Reliability: (E) = 0.86, (P) = 0.88, Responsiveness: (E) = 0.68, (P) = 0.85, Assurance: (E) = 0.87, (P) = 0.80) and Empathy: (E) = 0.83, (P) = 0.88] and they are considered to be good to excellent. This result was a proof of the questionnaire’s internal consistency. The level of statistical significance was set to α = 0.05.