Search PubMed⌕ Search
Home/Archive/PubMed

PubMed · 9192446

Predictive variable selection for the multivariate linear model.

J G Ibrahim·M H Chen
View original source ↗Export metadata ↓

Abstract

We develop a predictive Bayesian approach to variable selection in the multivariate linear model. A criterion derived from the Bayesian predictive density is proposed and a calibration is provided for it. Reference and informative priors are discussed, and an automated method that focuses on the response variable is proposed for specifying informative priors for the regression parameters. Relationships between the proposed criterion and other several well-known criteria are examined. Illustrative examples involving real data are given to demonstrate the methodology.

Explore related subjects

AsbestosBayes TheoremBiometryForced Expiratory VolumeHumansLinear ModelsLungOccupational ExposureVital CapacityJournal ArticleResearch Support, U.S. Gov't, P.H.S.

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

J G Ibrahim, M H Chen. 1997. Predictive variable selection for the multivariate linear model.. https://pubmed.ncbi.nlm.nih.gov/9192446/

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed1999

Case-control study of mesothelioma in South Africa.

BACKGROUND: South Africa has, uniquely, mined, transported, and used crocidolite, amosite, and chrysotile. A multicenter case-control study was done in South Africa to examine the details of asbestos exposure in cases and controls, and to calculate relative risks for level of certainty of asbestos exposure, nature of exposure (e.g., environmental, occupational) and fiber type. METHODS: Cases and controls (one cancer and one medical per case) were collected by six study centers from referral hospitals, and exposure information was collected by interviewing cases and controls in life. RESULTS: One hundred and twenty-three cases were accepted into the study. None had purely chrysotile exposure. Twenty-three cases had mined Cape crocidolite; three had mined amosite; and three Transvaal crocidolite plus amosite. A minimum of 22 of the cases had exclusively environmental exposure, 20 were from the NW Cape crocidolite mining area. The relative risks associated with environmental exposure in the NW Cape (crocidolite) were larger than for environmental exposure in the NE Transvaal (amosite and crocidolite): 21.9 vs. 7.1 and 50.9 vs. 12.0 for the cancer control and medical control datasets, respectively. CONCLUSIONS: The results confirm the importance of environmental exposure in the Cape crocidolite mining area, the relative paucity of cases linked to amosite, the rarity of chrysotile cases and are consistent with a fiber gradient in mesotheliomagenic potential for South African asbestos with crocidolite > amosite > chrysotile.

Asbestos↗
PubMed1999

Mortality among workers in the geothermal power plants at Larderello, Italy.

BACKGROUND: Since the early 1990s various European electricity companies have set up a research program, named EURELEX, devoted to the development of a comprehensive and validated title-occupation--job-exposure matrix (i.e., an instrument to assess exposures from occupations and jobs), in order to estimate and quantify past exposure to a number of potentially carcinogenic agents plus a few selected other toxic substances. The EURELEX program was first applied in Italy to a cohort of geothermal workers in Larderello, Tuscany. METHODS: The cohort mortality study comprised 4,237 men who had worked at the geothermal power plant between 1950 and 1990. Vital status and death certificates were obtained from registration offices in the municipality of birth or death. Particular attention was paid to neoplastic effects of exposure to asbestos, largely used in the past for pipeline insulating purposes. The matrix allowed the workers to be subdivided into several categories according to estimated absestos exposure. Smoking information was not available. RESULTS: No excess mortality was found for all causes and total cancer mortality. A small excess of pleural cancer (2 obs. vs. 1.4 exp.) was not statistically significant. The effect of other agents included in the matrix was also examined: the small excess of leukemia among those exposed to solvents prior to 1963 (year of benzene ban in Italy) did not reach the level of statistical significance. No cases of leukemia were found among the workers presumed to have been exposed to important ELF electromagnetic fields. CONCLUSION: The study provides reassuring information on a large cohort of a unique geothermal power plant since no significant excess mortality for asbestos related cancers was observed. Furthermore, it represents a useful application and validation of a comprehensive job-exposure matrix for the electric industry in Europe.

Asbestos↗