it would be good to have a null distribution (e.g. a marker that we would not expect is affected by a biomarker). we would compare our distributions to this null distribution.
one way to do this is to permute our data. imagine a grid overlaid on top of our data. any grid square with a tumor block will remain fixed. but all other grid squares can move randomly.
another thing to do is to plot different regions of the whole slide image and plot the CDFs for each region. this will allow us to see variance in the CDFs. if there is a clear difference between the M+ and M- plots, then we are great!
it would be good to have a null distribution (e.g. a marker that we would not expect is affected by a biomarker). we would compare our distributions to this null distribution.
one way to do this is to permute our data. imagine a grid overlaid on top of our data. any grid square with a tumor block will remain fixed. but all other grid squares can move randomly.
another thing to do is to plot different regions of the whole slide image and plot the CDFs for each region. this will allow us to see variance in the CDFs. if there is a clear difference between the M+ and M- plots, then we are great!