Question: Error Propagation--use of covariance
0
Kelsey Tyssowski10 wrote:

Hi! I am trying to determine the correct way to propagate my error.

I have a matrix (m) of expression values, and I am dividing the whole matrix by the first n=column(m/m[,1]). Then I am plotting the column means of my new matrix (n)--i.e. one point for each column.

I want to make sure I use the correct error bars on my plot. I know I need to propagate the error from dividing each column of m by m[,1] (I am using formulas here https://en.wikipedia.org/wiki/Propagation_of_uncertainty). I am confused about what I am supposed to do about the covariance. Should I use this term? If I use covariance, then the error on my plot for the first point (mean(m[,1]/m[,1])) is 0. It seems like maybe there should be some error represented here because there is error in m[,1]. On the other hand, I guess there is no error in n, because all of the points =1. I'm just not sure what the correct thing to do for plotting is.

Thanks!