Tests for equality of several alpha coefficients when their sample estimates are dependent

Abstract

coefficient alpha, test reliability, internal consistency, hypothesis testing, repeated measures,

Evaluating and Refining the Construct of Sexual Quality With Item Response Theory: Development of the Quality of Sex Inventory

“…Thus, these results help to validate the use of these scales across a wide range of sample diversity. However, comparable omnibus (Woodruff & Feldt, 1986) and pairwise comparisons (Abd-El-Fattah & Hassan, 2011)ofthese alpha coefficients suggested that the 12-item QSI-satisfaction subscale offered higher levels of internal consistency than the existing satisfaction scales in 18 of the 19 subgroups examined. Similarly, even the 6-item QSI-satisfaction subscale offered higher internal consistency than the existing satisfaction scales in 13 of the 19 subgroups, despite being markedly shorter in length.…”

“…As seen in Table 3, analyses in the 696 respondents providing 2-month follow-up data in the current project provided similar results. An omnibus test (Woodruff & Feldt, 1986) for differences across the six test-retest correlations of the satisfaction scales revealed a significant difference, v 2 (5) = 111.5, p\.0001, and subsequent pairwise comparisons (Abd-El-Fattah & Hassan, 2011) revealed that the 12-item QSI-satisfaction subscale had higher levels of test-retest reliability than the existing scales, despite being much shorter in length. Similar analyses in the 419 respondents reporting little to no change in their sexual satisfaction over time yielded stronger test-retest correlations for all of the scales (the follow-up survey did not include comparable questions about change in sexual dissatisfaction, and so it was not possible to calculate similar statistics for the QSI-dissatisfaction scales).…”

“…By carefully modeling each item's performance at different levels of satisfaction, the IRT analyses revealed dramatic differences in the precision and power of measurement offered by the different satisfaction scales, providing compelling evidence of the advantages of using the QSI scales. Differences in test-retest correlations and Cronbach's alphas among the six sexual satisfaction scales (the ISS, YSSS, PSSI, GMSEX, and QSI 12-and 6-item satisfaction scales) were first evaluated with an omnibus test recommended in Woodruff and Feldt (1986). These omnibus tests were all significant at p\.0001 (with v 2 (5)'s ranging from 53 to 3981).…”

Optimal number of response categories in rating scales: reliability, validity, discriminating power, and respondent preferences

“…Little further increase in reliability was found above seven response categories: alpha coefficients for scales with 7, 8, 9, 10, 11, and 101 response categories were all very similar. Using Feldt's test for the significance of differences between more than two related alpha coefficients (Woodruff & Feldt, 1986), it was determined that none of the differences between the alpha coefficients were statistically significant: using either of the recommended statistics HAN2 or UX1, χ 2 (10) < 18.31, p > .05.…”

Reducing Racial/Ethnic Disparities in Diabetes: The Coached Care (R2D2C2) Project

“…Comparisons of ethnic differences in the reliability coefficients of derived variables were performed using Feldt's W statistic (data not shown). 66 Differences between racial/ethnic groups on quality-ofcare, and all patient-reported measures were tested with separate linear regression models for continuous variables and logistic regression models for categorical variables declaring non-Hispanic white patients as the reference group. To account for multiple comparisons, group differences with p values≤0.001 were reported as statistically significant.…”