Yow-Jen Jou

Missing data treatment on travel time estimation for ATIS

This study proposes a missing data recovery method based on grey-relational nearest-neighbor substitution techniques for treating with missing data from dual-loop detectors in estimating travel time and evaluates the effects of the missing data on travel-time estimation performance. Field data from the Taiwan national freeway no.1 were used as a case study for testing the proposed model. Study results shown that the travel time estimation with missing data rate up to 33%. It is indicated that the proposed missing data treatment model can ensure the accuracy of travel time estimation with incomplete data sets.

A new multicollinearity diagnostic for generalized linear models

ABSTRACT We propose a new collinearity diagnostic tool for generalized linear models. The new diagnostic tool is termed the weighted variance inflation factor (WVIF) behaving exactly the same as the traditional variance inflation factor in the context of regression diagnostic, given data matrix normalized. Compared to the use of condition number (CN), WVIF shows more reliable information on how severe the situation is, when data collinearity does exist. An alternative estimator, a by-product of the new diagnostic, outperforms the ridge estimator in the presence of data collinearity in both aspects of WVIF and CN. Evidences are given through analyzing various real-world numerical examples.

Estimation of dynamic origin-destination by Gaussian state space model with unknown transition matrix

The dynamic origin-destination (O-D) pattern representing time-dependent trip demands from one place (origin) to another (destination) is amongst the most essential input data for most traffic operational analyses. Historical studies assumed that the transition matrix is known or at least approximately known, which is unrealistic for a real world network. Due to the fact that the number of trips to a specific destination, y, is easy to obtain and the O-D variable, x (path flow based in this research), is not directly observable, a Gaussian state space model is formulated to describe the relationships of x and y, observation equations, and the dynamics of x , state equations with unknown transition matrix. Under the assumption of Gaussian noise terms in the state space model, the distribution of the random transition matrix F is derived. A solution algorithm combining a Gibbs sampler and Kalman filter to tackle the problem of simultaneous estimation of F and x/sub t/ based on the latest available information is proposed. Real O-D data from the the Taipei rapid transit system is used to verify the presented model and solution method. Preliminary results are generally satisfactory, showing that in the unknown transition matrix case, significant estimates are also achieved.

VIF-based adaptive matrix perturbation method for heteroskedasticity-robust covariance estimators in the presence of multicollinearity

ABSTRACT In this study, we investigate linear regression having both heteroskedasticity and collinearity problems. We discuss the properties related to the perturbation method. Important observations are summarized as theorems. We then prove the main result that states the heteroskedasticity-robust variances can be improved and that the resulting bias is minimized by using the matrix perturbation method. We analyze a practical example for validation of the method.

Difference-based matrix perturbation method for semi-parametric regression with multicollinearity

ABSTRACT This paper addresses the collinearity problems in semi-parametric linear models. Under the difference-based settings, we introduce a new diagnostic, the difference-based variance inflation factor (DVIF), for detecting the presence of multicollinearity in semi-parametric models. The DVIF is then used to device a difference-based matrix perturbation method for solving the problem. The electricities distribution data set is analyzed, and numerical evidences validate the effectiveness of the proposed method.