I have data on protein levels from several hundred tumour tissue samples, which I am using in longitudinal survival models (Cox). Each experiment quantifies one protein, and is done in 20 batches each containing 120 samples. Batch x from any experiment contains the same tumours. However, I can see that *some* of my experiments show strong artefactual associations with batch. If I'm using one protein in a survival model, I can include batch as a covariate, or stratify by batch in stcox, which both ameliorate the batch effect.
The problem comes when I want to generate Cox models including several variables which are affected by batch in different ways. I feel there should be a way to correct my data for batch, but really don't know how to go about it. These corrected values would also be useful as input for other methods which do not allow for the inclusion of batch as a covariate.
What might you advise?
The problem comes when I want to generate Cox models including several variables which are affected by batch in different ways. I feel there should be a way to correct my data for batch, but really don't know how to go about it. These corrected values would also be useful as input for other methods which do not allow for the inclusion of batch as a covariate.
What might you advise?
Comment