paperKB
coga / coga-kb
Help
Sign in

Chunk #36 — OVERVIEW OF STATISTICAL METHODS — 2. Analysis of Incident Detection and/or Persistence of HPV — Frailty Models

Source
Marginal and mixed-effects models in the analysis of human papillomavirus natural history data.
Embedded
yes

Text

An alternative to the WLW method is to use frailty models to analyze multiple incident HPV infections(27). A frailty model introduces a frailty term, νi, to the Cox model as follows (1.10)λk(tik)=λ0(tik)vieθk+δzi(tik)+ηWi(tik) Where νi, treated as a random effect, represents the ith woman’s unobserved individual susceptibility to develop an oncogenic HPV infection relative to the average level in the population. The coefficient δ is interpreted as the log of the hazard ratio for any incident oncogenic HPV detection in a woman at a certain CD4+ / HIV RNA stratum compared with her risk if she were HIV-negative while holding all other factors constant. Frailty models are a form of mixed effects model, and the parameter estimate δ has a subject-specific interpretation. Model (1.10) can also be extended to assess HPV type-specific associations (not shown).