MODELING TRANSITION PROBABILITIES AND DURATION TIME FOR EVENT HISTORY DATA

• Yamashita Natsumi

  Department of Medical Informatics, Kyushu University Hospital

• Nagata Eiko

  Biostatistics Center, Kurume University

• Ohyama Tetsuji

  General Clinical Research Center, Oita University Hospital

• Yanagawa Takashi

  Biostatistics Center, Kurume University

Abstract

There could be covariates that are related to transition probabilities, but not related to duration times to the next event, or vice versa, in event history analyses. A method is proposed in this paper that estimates those relationships, separately, by using an accelerated failure time model. The method is applied to the data taken from individuals staying at home and having government care service, where different levels of care are provided to care need individuals based on their severities officially recognized. It is shown among others that family and disease are significantly related to duration times to the next event, but not related to transition probabilities.

Journal

• Bulletin of informatics and cybernetics

  Bulletin of informatics and cybernetics 42 71-79, 2010-12

  Research Association of Statistical Sciences

Keywords

• accelerated failure time model
• event history analysis
• Markov process
• survival analysis

Details

• CRID: 1390290699825853696
• NII Article ID: 120005208806
• NII Book ID: AA10634475
• DOI: 10.5109/25907
• ISSN: 2435743X
• ISSN: 0286522X
• HANDLE: 2324/25907
• Text Lang: en
• Data Source:
  • JaLC
  • IRDB
  • Crossref
  • CiNii Articles
• Abstract License Flag: Allowed