The impact of data censorship on pre-movement time prediction in building fire evacuation: focusing on people with learning disabilities
Year: 2025
Type: article
Abstract: In fire evacuations, pre-movement time (PT) is an important part of the total egress time. Data from fire drills is commonly used to investigate occupant behavior, including PT, but these data are often subjected to censoring. This study examines the impact of data censoring on PT analysis during fire evacuation drills, which is important for predicting overall evacuation time. It identifies various types of data censorship and highlights the importance of accounting for censored data in statistical analyses. The research utilizes the Accelerated Failure Time model (AFT), a parametric survival analysis method, to address data censoring issues. Data from two fire evacuation drills in residential buildings with occupants with learning disability were analyzed using two scenarios: one excluding censored data and another incorporating it. The results demonstrate that accounting for censored data significantly alters the interpretation of the causal model, with covariates such as fire incident time emerging as significant only when censored data is considered. The AFT model effectively manages censored data by updating predictions for observed events and extrapolating values for censored observations using the best-fit distribution. This study underscores the importance of incorporating censored data into evacuation models emphasizing that excessive data censoring reduces the validity of AFT model predictions. Although limited by sample size, these findings offer valuable insights into effective covariates influencing PT and provide guidance for future evacuation modeling and simulation tools.
Source: Safety Science
Language: English
Subfield: Ocean Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,360
Awards: G1442795146, G4412656742
OpenAlex ID: W7117371883