Thank you Dr. Smith and Dr. Hillis for providing us with MRMCaov package.
I have a planned analysis from an MRMC study where we only have binary ratings for the presence or absence of lesions from readers using two imaging modalities (let's say modality A and modality B). Since there are no confidence scores from readers, what we can do is to find if sensitivity (or specificity) improves in modality B as compared to modality A. I understand that this can be done by utilizing the following code paradigm in MRMCaov.
mrmc(
binary_sens(truth, binary_rating), treatment, reader, case, data = VanDyke
)
However, there are some challenges because the the data is clustered. That is, one image can have many lesions (on an average 2-3 lesions per images). I haven't found any documentation that says MRMCaov handles clustered ROC analysis. Can you please let me know about this?
I have been exploring other options such OBUMRM2 fortran code, but it seems like even though this can handle clustered data, it can only be used for AUC analysis and not for sensitivity/specificity analysis.
Thanks in advance.
Thank you Dr. Smith and Dr. Hillis for providing us with MRMCaov package.
I have a planned analysis from an MRMC study where we only have binary ratings for the presence or absence of lesions from readers using two imaging modalities (let's say modality A and modality B). Since there are no confidence scores from readers, what we can do is to find if sensitivity (or specificity) improves in modality B as compared to modality A. I understand that this can be done by utilizing the following code paradigm in MRMCaov.
However, there are some challenges because the the data is clustered. That is, one image can have many lesions (on an average 2-3 lesions per images). I haven't found any documentation that says MRMCaov handles clustered ROC analysis. Can you please let me know about this?
I have been exploring other options such OBUMRM2 fortran code, but it seems like even though this can handle clustered data, it can only be used for AUC analysis and not for sensitivity/specificity analysis.
Thanks in advance.