Skip to main content

Table 3 Standard error in log scale with and without model misspecification (n = 1000)

From: Comparing performance between log-binomial and robust Poisson regression models for estimating risk ratios under model misspecification

  1. Unshaded: Models were correctly specified
  2. Light shaded: Misspecified linear predictors or misspecified link function
  3. Dark shaded: Misspecified linear predictors and misspecified link function
  4. LB: log-binomial, RP: robust Poisson
  5. Change of scenarios: Increasing intercept: I → II → III; Increasing coefficient of β2: IV → III → V; Change of link function: III (log), VI (logit), VII (probit)
Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us