2

I am interested in plotting correlation heat maps for a multivariate dataset collected in two batches. Some of the variables show a clear batch effect others do not show any batch effect. Is there any way to take care of batch effects in correlations?

Karolis Koncevičius
  • 4,282
  • 7
  • 30
  • 47
statstar
  • 61
  • 1
  • 4
  • 3
    What exactly do you mean by "batch effect in correlations", and what would constitute having 'taken care of them'? – gung - Reinstate Monica Mar 05 '13 at 01:20
  • 1
    I have some clinical and omics data variables.Together there are 137 variables.Data is collected for mothers as well as their babies. Patients were recruited and omics was done in two batches.There is clearly difference between absolute values in two different batches in some of the variables.The idea is to look at correlations between these variables first for exploratory purposes.What is a good way to normalize such a data for batch effects.If we merge the data without normalization would it not lead to spurious results . – statstar Mar 05 '13 at 20:55

1 Answers1

1

You have only two batches, what about making side-by-side two heatmaps, taking care of color coding (and, if used, reordering of variables) is identical in the two plots?

The part of the question about normalization I leave for others.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467