I’m running a DIABLO analysis with several data integration combinations. But I’ve a theorical question, the consensus plot represented below, it indicates the discrimination of samples only if you use for the prediction scheme “average prediction’ scheme” (averanging all scores from each component of each dataset)?
Is there a consensus plot if you’re using majority vote or weighted vote scheme?
https://bitbucket-assetroot.s3.amazonaws.com/repository/qMzjR4/3883906484-2018-10-08_145730.png?Signature=D%2BiVxVwHQD58Jcly5G5rp%2BNMa28%3D&Expires=1560911790&AWSAccessKeyId=AKIAIQWXW6WLXMB5QZAQ
Hi Natalia,
Thanks for your enquiry.
The voting scheme is for prediction of new samples so that the disagreement of studies can be reconciled, while in consensus sample plots - as mentioned in the section 2.4 of the paper - the component averaging does not consider weights. Is that something you would find useful? And if so, what do you propose to use as weights for each sample?
Best,
Al
So, this consensus plot only represents samples classification if you use the voting scheme
" average prediction", wright?
@nataliabueno you are correct, the current consensus plot indicates the discrimination of samples based on the average scheme only. This is not possible with the majority vote schemes as we do not have “consensus component” scores (we use the predicted class labels directly to compute the majority vote class).