Hi, I'm fairly new to R studio and have a basic grasp on statistics which is why I need some guidence. Please redirect me to the right category if I'm posting in the wrong place.
Im trying to estimate a cut off with a 95% CI for a biomarker with Gaussian Mixture Modeling. The dataset is biomdal and the cut-off is determined as the crossing point of the right tail of the first fit and the left tail of the second fit with bootstrapping.
The problem is that the datset is not all that tidy, with extreme values on both ends and the two distributions are not perfectly normal. I don't have a good reason to exclude these extreme values (which fixes the cut-off estimation) but they completley throw the CI intervals out of whack (e.g. extending the upper limit all the way to the end of the second distribution). I tried bootsrapping from the fits instead of raw values which also fixes the unstable CI issue but I'm not sure if this is statistically okay to do. I'm also thinking if the problem lies in the R code and methodology rather than a difficult dataset. Any ideas on how to get around this?