### Abstract

esƟmated using a Markov Chain Monte Carlo simulation method. From the model outputs smooth trend esƟmates can be computed at various aggregation levels for the mean number of trip legs per person per day and the mean distance traveled per trip leg, as well as for derived quantities such as the mean distance per person per day. We discuss the model building and evaluation processes as well as the results based on the fitted models.

Original language | English |
---|---|

Publisher | Statistics Netherlands |

Number of pages | 65 |

Publication status | Published - 27 Sep 2019 |

### Cite this

*Multilevel time-series modeling of mobility trends - Final Report*. Statistics Netherlands.

}

**Multilevel time-series modeling of mobility trends - Final Report.** / Boonstra, Harm Jan ; van den Brakel, Jan; Das, Sumon.

Research output: Working paper › Discussion paper › Professional

TY - UNPB

T1 - Multilevel time-series modeling of mobility trends - Final Report

AU - Boonstra, Harm Jan

AU - van den Brakel, Jan

AU - Das, Sumon

N1 - Dutch Travel Survey microdata for the period of 1999-2017 have been used for calculating direct estimates and their standard errors, which are then used as input in time series multilevel model building.

PY - 2019/9/27

Y1 - 2019/9/27

N2 - This report describes the time series models developed for the mobility trend estimation project carried out by Statistics Netherlands in collaboration with KiM/Rijkswaterstaat. First, direct estimates along with standard error estimates are obtained for each year in the period 1999-2017 from the microdata of the Dutch Travel Survey for a detailed cross-classification by person characteristics sex and age class and trip leg characteristics mode and purpose. Consequently, these direct estimates are smoothed by modeling them using multi-level time series models that account for influential outliers as well as for the redesigns of the survey within the Ɵme span considered. Two target variables are modeled in this way: the number of trip legs per person per day and the distance traveled per trip leg. The models are specified in a hierarchical Bayesian framework andesƟmated using a Markov Chain Monte Carlo simulation method. From the model outputs smooth trend esƟmates can be computed at various aggregation levels for the mean number of trip legs per person per day and the mean distance traveled per trip leg, as well as for derived quantities such as the mean distance per person per day. We discuss the model building and evaluation processes as well as the results based on the fitted models.

AB - This report describes the time series models developed for the mobility trend estimation project carried out by Statistics Netherlands in collaboration with KiM/Rijkswaterstaat. First, direct estimates along with standard error estimates are obtained for each year in the period 1999-2017 from the microdata of the Dutch Travel Survey for a detailed cross-classification by person characteristics sex and age class and trip leg characteristics mode and purpose. Consequently, these direct estimates are smoothed by modeling them using multi-level time series models that account for influential outliers as well as for the redesigns of the survey within the Ɵme span considered. Two target variables are modeled in this way: the number of trip legs per person per day and the distance traveled per trip leg. The models are specified in a hierarchical Bayesian framework andesƟmated using a Markov Chain Monte Carlo simulation method. From the model outputs smooth trend esƟmates can be computed at various aggregation levels for the mean number of trip legs per person per day and the mean distance traveled per trip leg, as well as for derived quantities such as the mean distance per person per day. We discuss the model building and evaluation processes as well as the results based on the fitted models.

M3 - Discussion paper

BT - Multilevel time-series modeling of mobility trends - Final Report

PB - Statistics Netherlands

ER -