China 
Africa 
Explore chapters and articles related to this topic
Several Other Distributions
Published in Kenneth G. Russell, Design of Experiments for Generalized Linear Models, 2018
Finally, it considers situations where the nature of the distribution of the response variable is unknown, but it is believed that the link function and variance function can be specified. Analyses in this situation are often called quasi-likelihood analyses.
Efficient Integration of Sufficient Dimension Reduction and Prediction in Discriminant Analysis
Published in Technometrics, 2019
The BIC procedure for selecting the structural dimension is often preferred over AIC in the dimension reduction literature (Cook and Forzani 2009, e.g.). Recently, Zhang and Mai (2018) showed that the BIC is asymptotically consistent for selecting the true envelope dimension under weak moment conditions in a model-free context. Under their quasi-likelihood framework, the ENDS objective function from (5.1) can be viewed as a partially optimized quasi-likelihood function. Therefore, we propose the following unified information criterion for selecting the ENDS dimension, where h(n) = 2 for AIC and log (n) for BIC. It is easy to verify that this unified information criterion reduces to the classical AIC and BIC of ENDS-LDA when λ = 1 and of ENDS-QDA when λ = 0. We demonstrate the AIC, BIC, and five-fold cross-validation in our numerical studies.
Map-matching poor-quality GPS data in urban environments: the pgMapMatch package
Published in Transportation Planning and Technology, 2019
Adam Millard-Ball, Robert C. Hampshire, Rachel R. Weinberger
Our approach uses a three-part quasi-likelihood function to rank candidate paths, comprising the following components: A geometric component, based on the distance between each vertex on the GPS trace and the corresponding vertex on each candidate pathA topological component, based on the ratio of segment lengths of the GPS trace and candidate path (a segment is defined as the path between two vertices on the GPS trace)A temporal component, based on the implied speed of each segment of the candidate path
The effect of international oil price on LNG price in South Korea and Japan
Published in Geosystem Engineering, 2018
Using this log-likelihood function, parameters both univariate GARCH (ϕ*) and correlation structure (ψ*) are estimated by maximizing log-likelihood function. To begin with, Rt is substituted by the identity matrix It to estimate the parameter set ϕ = (ϕ1, ⋯, ϕn), resulting in quasi-likelihood function: