1
Friends, My doubt is as follows, I am using the following code in R to construct a ROC curve , but I cannot say if in case of GLMM (Mixed Generalized Linear Model) I can use and if in this case the code then will consider the random part as well or only the fixed ones as occurs in the Logistic Regression, which would not give me correct information.
library(ROCR)
pred <- prediction(fitted(gg1), dados$status)
perf <- performance(pred, measure = "tpr", x.measure = "fpr")
plot(perf, main = "ROC", col = "blue", lwd = 3)
abline(a = 0, b = 1, lwd = 2, lty = 2)
perf.auc <- performance(pred, measure = "auc")
(area <- performance(pred, "auc"))
The ROC curve simply categorizes a continuous variable in a 2-category category category from another categorical variable (https://gpestatistica.netlify.com/blog/curvaroc/). If the
fitted(gg1)consider the random effect, then the curve "will also go". Only it will depend on the shape you specify the mixed model. Is your dice like before and after? Using a ROC curve for Before, and another for After would not be more interesting? (because the cutoff points should be quite different for these 2 moments).– Guilherme Parreira