China
Africa
Explore chapters and articles related to this topic
Back-propagation neural networks and generalized linear mixed models to investigate vehicular flow and weather data relationships with crash severity in urban road segments
Published in Gianluca Dell’Acqua, Fred Wegman, Transport Infrastructure and Systems, 2017
L. Mussone, M. Bassani, P. Masci
The GLMM model is a regression model of a response variable that contains both fixed and random effects and comprises data, a model description, fitted coefficients, co-variance parameters, design matrices, residuals, residual plots, and other diagnostic information. Fixed-effects terms usually refer to the conventional linear regression part of the model. Random effects terms are associated with individual experimental units taken at random from a population, and account for variations between groups that might affect the response. The random effects have prior distributions, whereas the fixed effects do not.
Optimizing the guiding sign system to improve drivers’ lane-changing behavior at freeway exit ramp
Published in Journal of Transportation Safety & Security, 2022
Wang Xiang, Yanjie He, Qunjie Peng, Xiaomeng Li, Qingwan Xue, Guiqiu Xu
The GLMM is a new-type statistical regression model developed based on the Generalized Linear Model (GLM) and the Linear Mixed Model (LMM) and composed of three components: a random component, a system component (i.e., fixed effect and random effect), and a connection function. As a regression model, GLMM adopts the Maximum Likelihood Estimation (MLE) method to fit discrete dependent repeatedly measured data, including imbalanced panel data. Drawing upon the idea of mixed model, GLMM introduces the random effect on the basis of linear prediction (Ando et al., 2022; Liu & Xiang, 2019). Thus, it can reflect the relevance between repeated measurements of dependent variables via the covariance matrix of the random effect. Meanwhile, the introduction of the random effect has addressed the issues of correlation between data, excessive discreteness, and heterogeneity (Buratin et al., 2022; Lee & Yoo, 2014; Papantoniou, 2015). Furthermore, the GLMM model has modified the requirement in previous statistical models that dependent variables obey a normal distribution by introducing the connection function to address the issue of non-normal distribution of dependent variables. Lastly, it is worth reiterating that this model has no specific constraint on the type of independent variables. It was often used in previous studies to probe into traffic related problems (Choudhary & Velaga, 2017, 2019; Yadav & Velaga, 2020). The significance level of the statistical tests was 0.05.
Evaluation of pavement surface texture at the network level
Published in Nondestructive Testing and Evaluation, 2019
Shuvo Islam, Mustaque Hossain, Richard Miller
Since a number of sections are from the same geographical region (district or county), there could be a spatial association among the observations from similar regions and could impart random effect on the analysis [11]. To account for such association, a generalised linear mixed model (GLMM) approach was used in this study. A GLMM approach enables statisticians to incorporate both fixed and random effects in a model [12]. Additionally, since the number of roadway sections that are available only for a particular surface type and not for other surface types, the design structure of the data essentially becomes nested where road sections are nested within surface types. In other words, levels of the factor road section are nested within the factor surface types and one or more observations on each road-section*surface type combination. Thus a two-way mixed effect nested model has been used in this study to access how much MTDs vary across different surface types of different surface ages. Equation 1 describes the model.
The short-term influence of rear wheel axle position and training on manual wheelchair propulsion technique in novice able-bodied participants during steady-state treadmill propulsion, a pilot study
Published in Assistive Technology, 2020
Ian Rice, Chandrasekaran Jayaraman, Ryan T. Pohlig
Descriptive statistics were calculated as estimated marginal means and standard errors (SEs) for continuous variables (Table 2), observed means and standard deviations in Tables 1 and 4, while descriptive data were presented as percentages in Table 3. The primary research hypotheses evaluating whether optimization and training affected technique were tested using GLMM. The mixed model was chosen because it offers advantages over a repeated measures analysis of variance (ANOVA). First, the GLMM procedure allows the direct specification of the residual covariance matrix, which negates the need to satisfy the restricting assumptions of homogeneity of covariance matrices and sphericity present in ANOVA. Second, GLMMs are able to include time-varying covariates, which classic ANOVA cannot. In this case, speed was included as a time-varying covariate because it is known that pushing at faster speeds affect the outcomes independent of training or optimization. Model fit was evaluated by comparing the Akaike information criterion (AIC) and Baysian information criterion (BIC; Posada & Buckley, 2004). Significant interaction effects were investigated by pairwise comparisons testing the simple main effect of time within each group. Marginal pairwise comparisons followed significant main effects. Fisher’s least significant difference (LSD) procedure was used for all follow-up tests. All assumptions were evaluated, and alpha was set at the nominal level of 0.05 for each outcome.