I have conducted a binary logistic regression.
There are 294 lesions and logistic regression can separate them as benign and malignant.
For this purpose three of the study parameters produced a formula. As you know it is something like:
Benign/Malign Score = 1/ 1+ exp[(Parameter1*Constant1)+(P2*C2)+(P3*C3)]
During cross validation, 80% of cases constitutes training set and %20 of them are test set.
Training and test set are formed randomly.
The constants in this formula change during each rebuild of the training and test groups as well as Sensitivity, Specificity and Accuracy.
I need help: 1- How would you suggest I manage the validation process so that I can give a single formula that everyone can use? Because the formula is changing in every single validation.
2- I want to study validation process 500 times. I am going to give arithmetic mean of Sensitivity, Specificity and Accuracy of 500 measurements. Do you offer another information to provide in the manuscript (for example min, max, 95CI of Accuracy etc.)?
Many thanks in advance to everyone who showed interest and kindness to respond.
In my opinion your work is closely related to a simulation study. rather than a cross-validation. You may benefit from the excellent remarks of Mary Davidian attached.
I hope this may help.
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training