China 
Africa 
Explore chapters and articles related to this topic
Structural Equation Modeling
Published in Jhareswar Maiti, Multivariate Statistical Modeling in Engineering and Management, 2023
A related issue, which is often overlooked by researchers, is the computation of covariance and correlation between two variables when one or both are non-metric in nature. Data can be metric (measured using ratio or interval scale) or non-metric (measured using ordinal or nominal scale). The Pearson product moment correlation is developed for metric data. The types of correlations between two variables, say X and Y are given below. Pearson product moment correlation: When both X and Y are measured using metric scale (ratio or interval).Tetrachoric correlation: When both X and Y are binary, i.e., each is measured with two categories (dichotomous).Polychoric correlation: When both X and Y are polytomous, i.e., each is measured with three or more categories.Biserial correlation: When one of X and Y is metric and another one is binary (dichotomous).Polyserial correlation: When one of X and Y is metric and another one is polytomous.
Dynamic Measures of Effectiveness in Command and Control
Published in Peter Berggren, Staffan NäHlinder, Erland Svensson, Assessing Command and Control Effectiveness, 2017
We have mentioned that discrete time series can, and often do, represent genuine continuous variables; repeated questions at an ordinal level of genuinely continuous psychological variables (for instance mental workload, situational awareness, performance) are examples of this. The general factor analytic model is adapted to correlations between underlying continuous manifest variables rather than to direct correlations between discrete and ordinal manifest variables. The most commonly used input matrix to factor analyses is based on Pearson’s product moment correlations. This covariation measure can, however, render suboptimal estimates of underlying correlations between continuous variables. The polychoric correlation is a better alternative as it represents the ‘true’ correlation between underlying continuous variables. Factor loading estimates based on polychoric correlations are found to be less biased and often higher than loadings based on product moment correlations. Accordingly, factors based on polychoric correlations are expected to be more reliable. The polychoric coefficient is estimated from a contingency table formed by observations on the two ordinal manifest variables (Jöreskog and Sörbom 1996; Zhang and Browne 2010). By means of a preprocessor to LISREL (PRELIS), multivariate data screening and estimations of polychoric, as well as Pearson correlations, can be made (Jöreskog and Sörbom 1996).
A comprehensive sequential strategy for structural equation modeling of traffic barrier crashes
Published in Journal of Transportation Safety & Security, 2021
Mahdi Rezapour, Shaun S. Wulff, Khaled Ksaibati
The recommended approach of Rosseel (2014) did not work in this study because it was not possible to constrain the corresponding variances to be positive, which led to improper or nonconvergent model fits. The chosen alternative was to directly use polychoric correlations, which are appropriate measures of the correlations for ordinal predictors and tetrachoric correlations, which are appropriate measures of the correlations between binary predictors. Tetrachoric correlations can be considered to be a special case of the polychoric correlations (Olsson, 1979). The direct use of these correlations limited the available estimation approaches. However, Unweighted Least Squares (ULS) provided good model fits in this study without the need to assume multivariate normality. ULS or uses (3) with or no weighting matrix. Furthermore, ULS was applied to the corresponding polychoric covariance matrix rather than the sample covariance matrix (Bollen, 1989). This amounts to minimizing the differences between the estimated covariance matrix and
Exploring COVID-19 pandemic potential impacts on students’ school travel behavior
Published in Transportation Letters, 2023
Mohamed Abouelela, Mohamed Samir, Constantinos Antoniou
We used polychoric correlation coefficients as the base for EFA calculation. The polychoric correlation has been proven to be more appropriate for ordinal data than Pearson’s correlation, which assumes fixed intervals between the different values, which is not the case for ordinal data (Holgado-Tello et al. 2010). Evaluation scores were ordered in ascending order, and the polychoric correlation matrix was calculated using the polycor package (Fox 2019), in R (R Core Team 2020). After calculating the correlation matrix, we calculated EFA using the psych package (Revelle 2019), also in R (R Core Team 2020), applying a varimax rotation.