Evaluating the effect of neighbourhood weight matrices on smoothing properties of Conditional Autoregressive (CAR) models

Earnest, Arul; Morgan, Geoff; Mengersen, Kerrie; Ryan, Louise; Summerhayes, Richard; Beard, John
January 2007
International Journal of Health Geographics;2007, Vol. 6, p54
Academic Journal
Background: The Conditional Autoregressive (CAR) model is widely used in many small-area ecological studies to analyse outcomes measured at an areal level. There has been little evaluation of the influence of different neighbourhood weight matrix structures on the amount of smoothing performed by the CAR model. We examined this issue in detail. Methods: We created several neighbourhood weight matrices and applied them to a large dataset of births and birth defects in New South Wales (NSW), Australia within 198 Statistical Local Areas. Between the years 1995-2003, there were 17,595 geocoded birth defects and 770,638 geocoded birth records with available data. Spatio-temporal models were developed with data from 1995-2000 and their fit evaluated within the following time period: 2001-2003. Results: We were able to create four adjacency-based weight matrices, seven distance-based weight matrices and one matrix based on similarity in terms of a key covariate (i.e. maternal age). In terms of agreement between observed and predicted relative risks, categorised in epidemiologically relevant groups, generally the distance-based matrices performed better than the adjacency-based neighbourhoods. In terms of recovering the underlying risk structure, the weight-7 model (smoothing by maternal-age 'Covariate model') was able to correctly classify 35/47 high-risk areas (sensitivity 74%) with a specificity of 47%, and the 'Gravity' model had sensitivity and specificity values of 74% and 39% respectively. Conclusion: We found considerable differences in the smoothing properties of the CAR model, depending on the type of neighbours specified. This in turn had an effect on the models' ability to recover the observed risk in an area. Prior to risk mapping or ecological modelling, an exploratory analysis of the neighbourhood weight matrix to guide the choice of a suitable weight matrix is recommended. Alternatively, the weight matrix can be chosen a priori based on decision-theoretic considerations including loss, cost and inferential aims.


Related Articles

  • Predictive value and completeness of the registration of congenital abnormalities in three Danish population-based registries. Larsen, Helle; Nielsen, Gunnar L.; Bendsen, Jørgen; Flint, Carolin; Olsen, Jørn; Sørensen, Henrik T. // Scandinavian Journal of Public Health;Jan2003, Vol. 31 Issue 1, p12 

    Aims: The predictive value and completeness of data on congenital abnormalities (CAs) collected in three administrative health registries in the County of North Jutland, Denmark were compared. Methods: The study included all singleton liveborn infants in the county during the period 1991-94...

  • Birth Defects Surveillance: Assessing the `Gold Standard'. Honein, Margaret A.; Paulozzi, Leonard J. // American Journal of Public Health;Aug1999, Vol. 89 Issue 8, p1238 

    Objectives. This study assessed the sensitivity of the Metropolitan Atlanta Congenital Defects Program (MACDP) by capitalizing on the delayed receipt of a data source. Methods. In 1997, we reviewed the medical records of potential cases from the 1995 birth certificates that had not previously...

  • The Surveillance of Birth Defects: The Usefulness of the Revised US Standard Birth Certificate. Watkins, Margaret L.; Edmonds, Larry; McCleam, Anne; Mullins, Lynda; Mulinare, Joseph; Khoury, Muin // American Journal of Public Health;May96, Vol. 86 Issue 5, p731 

    To assess the sensitivity and positive predictive value of birth defects reported on the 1989 revision of the US Standard Birth Certificate, a population of 76 862 Atlanta-area births during 1989 and 1990 was used as the basis for comparing 771 birth certificates that reported birth defects with...

  • A Microcomputer-Based Vital Records Data Base with Interactive Graphic Assessment for States and Localities. Wartenberg, Daniel; Agamennone, Vincent J.; Ozonoff, David; Berry, R. J. // American Journal of Public Health;Nov89, Vol. 79 Issue 11, p1531 

    Abstract: Vital records data bases describe large populations over long periods of time, yet their organization and size often preclude or discourage their use. We constructed a microcomputer-based data base of all singleton births in Massachusetts, 1975-84. The original data were stored in...

  • Control Selection and Participation in an Ongoing, Population-based, Case-Control Study of Birth Defects. Cogswell, Mary E.; Bitsko, Rebecca H.; Anderka, Marlene; Caton, Alissa R.; Feldkamp, Marcia L.; Sherlock, Stacey M. Hockett; Meyer, Robert E.; Ramadhani, Tunu; Robbins, James M.; Shaw, Gary M.; Mathews, T. J.; Royle, Marjorie; Reefhuis, Jennita // American Journal of Epidemiology;Oct2009, Vol. 170 Issue 8, p975 

    To evaluate the representativeness of controls in an ongoing, population-based, case-control study of birth defects in 10 centers across the United States, researchers compared 1997–2003 birth certificate data linked to selected controls (n = 6,681) and control participants (n = 4,395)...

  • Deaths in 121 U.S. cities, week ending February 25, 1995 (8th week).  // MMWR: Morbidity & Mortality Weekly Report;3/3/1995, Vol. 44 Issue 8, p148 

    Presents a chart which details the number of reported deaths in 121 cities of the United States for the week ending February 25, 1995. Reporting areas; Totals.

  • Handwriting Example XIII.  // Swedish American Genealogist;Mar2007, Vol. 27 Issue 1, p28 

    The article presents an image of handwritten birth records from Nordmark parish in eastern Värmland, Sweden.

  • A method of creating a death register for general practice. Khunti, Kamlesh // BMJ: British Medical Journal (International Edition);4/13/96, Vol. 312 Issue 7036, p952 

    Describes a system of creating an accurate death register for general practitioners. Audits of death register; Use of preventing notes returned to the family health services authority without the cause of death being recorded on the computer.

  • Are causes of death predictable? Tallis, G.M.; Leppard, P.; Tallis, G.A.; Hansen, D. // Human Biology;Feb98, Vol. 70 Issue 1, p117 

    Looks at the selection of sample death records of adult males from 1967 to 1970 from the South Australian Registry of Births, Deaths and Marriages. Information on subjects and methodologies; Further detailed information.


Read the Article


Sorry, but this item is not currently available from your library.

Try another library?
Sign out of this library

Other Topics