mis is a
link-glm object that specifies the link function in Neuhaus (1999, expression~(8)) for handling misclassified responses in binomial regression models using maximum likelihood. A prior specification of the sensitivity and specificity is required.
mis(link = "logit", sensitivity = 1, specificity = 1)
the baseline link to be used.
the probability of observing a success given that a success actually took place given any covariate values.
the probability of observing a failure given that a failure actually took place given any covariate values.
specificity should be greater or equal
to 1, otherwise it is implied that the procedure producing the
responses performs worse than chance in terms of misclassification.
Neuhaus J M (1999). Bias and efficiency loss due to misclassified responses in binary regression. Biometrika, **86**, 843-855 https://www.jstor.org/stable/2673589
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
## Define a few links with some misclassification logit_mis <- mis(link = "logit", sensitivity = 0.9, specificity = 0.9) lizards_f <- cbind(grahami, opalinus) ~ height + diameter + light + time lizardsML <- glm(lizards_f, family = binomial(logit), data = lizards) lizardsML_mis <- update(lizardsML, family = binomial(logit_mis), start = coef(lizardsML)) ## A notable change is coefficients is noted here compared to when ## specificity and sensitity are 1 coef(lizardsML) coef(lizardsML_mis) ## Bias reduction is also possible update(lizardsML_mis, method = "brglmFit", type = "AS_mean", start = coef(lizardsML)) update(lizardsML_mis, method = "brglmFit", type = "AS_median", start = coef(lizardsML))
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.