I am going to plot an overfitting- corrected calibration curve(using the cross-validation method) for a regression model with Poisson family (link function=log) in R.
I know that this can be done for the logistic, ols, and cox modelusing rms package.
Does anyone know about the R codes for doing overfitting- corrected calibration curve of Poisson model?
Thanks in advance
Hello
Before using the Poisson regression you must verify the assumptions that must be fulfilled, otherwise, the application of the Poisson regression is not valid (Mean and variance must be equal)
How can you "correct" for overfitting? As far as I can imagine, the best ways to avoid overfitting are:
* use a model function that is appropriate for the data;
* reduce the number of parameters to fit;
* use as much as possible data points.
What are you calibrating? Could you share some data with us?
Dear Koen
Actually, I'm trying to get a curve like the image below.
In order to bias-corrected (or overfitting-corrected) curve, a 10-fold cross-validation method is considered.
I need this curve for modified Poisson regression.
Shadi Naderyan Fe'li I would like to see your RAW data points and some information: what did you measure exactly and what kind of relationship do you expect?
- Badges
- Science topic
- Similar topics
- Condensed Matter Physics
- Zero