For correlation=TRUE about mixrf HOT 3 CLOSED

The idea is to use cross-validation to estimate the correlation. get.CVlist() will split the data into nCV fold and return the list of CV samples. av.samp is the available sample index with observations, and this depends on your data, not only the number of patients/tissues. When the length of av.samp cannot evenly be divided by nCV. split() will throw a warning, but this only means that the sample split is not even, and only differs by one, which is unavoidable.

from mixrf.

The idea is to use cross-validation to estimate the correlation. get.CVlist() will split the data into nCV fold and return the list of CV samples. av.samp is the available sample index with observations, and this depends on your data, not only the number of patients/tissues. When the length of av.samp cannot evenly be divided by nCV. split() will throw a warning, but this only means that the sample split is not even, and only differs by one, which is unavoidable.

Hi Jiebiao,
Thanks for your fast and detailed response~

Is it easier to debug it on your end to make it more adaptive? I hope only differs by one do not have to throw error. I am indeed not an R expert and after dozens' of hours of failure (I should initially try cross-validation on baby dataset first), I might start debugging this issue next week.

Comparatively, for fitted result, I used to run even 12-24 hours and never confront any problem.

Best,
Yezheng

from mixrf.

My understanding is that this is just a warning, not an error. It would not affect the results other than the uneven splitting of available samples in cross-validation. If you do encounter an error, please share a reproducible code example.

from mixrf.