Purposeful selection of variables in logistic regression.
Bursac Z, Gauss CH, Williams DK, Hosmer DW. · 2008
Read Purposeful selection of variables in logistic regression. by Bursac Z, Gauss CH, Williams DK, Hosmer DW. (2008) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc2633005.txt.
Description
Bursac, Gauss, Williams, and Hosmer automated a model-building strategy for logistic regression in 2008, the statistical workhorse for yes-or-no medical outcomes. Choosing which candidate variables belong in a model is the central problem, and mechanical selection algorithms have known limitations. Statisticians Hosmer and Lemeshow had described purposeful selection. There, the analyst judges each variable step by step. This paper turns that craft into an algorithm.
The team tested their macro against three standard procedures in SAS software, forward, backward, and stepwise selection, in simulations. The advantage appeared when the goal was risk factor modelling rather than pure prediction. Beyond clearly significant variables, the method keeps important confounding variables. Mechanical rules drop those. The result is a slightly richer model. They demonstrate it on the Worcester Heart Attack Study data.
Catalog record
| CREATOR | Bursac Z, Gauss CH, Williams DK, Hosmer DW. |
|---|---|
| YEAR | 2008 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 25 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |