I’m afraid this may be a very stupid question and far more general than cellprofiler, but anyway…
For each biological replicate I have several technical replicates. How do I calculate an error (or the SD for that matter) for the biological replicates that takes into account the errors of the replicates?
Tthey way this relates to cellprofiler is that I’m actually considering the variability of a nuclear signal in nuclei in a single well (96-well plate) when the result will be the average of three wells.
Thanks very much!
Are you looking for just general information on error propagation?
I’m sorry I didn’t reply earlier.
It isn’t error propagation (or, well, in a way it is, but rather trivial as it’s all averages).
My question is rather silly, I think:
I have 3 wells in a plate.
From each well I take 9 pictures.
Each pictures has some 40 cells I can quantify.
Now, I want to determine the well-to-well variation (say, a CV). Do I average all the cells (cell measurements) in a well, then use that average as a mean for each well? If so, the CV (or SEM) among the 3 wells won’t take into account the SEM for the mean that describes each well.
That is, if I use the mean for each well, then from that I calculate the well-to-well variation (CV, or SEM for the mean of 3 wells), intuitevly I’m not taking into account the uncertainty that is already in the calculation of each mean.
Thanks very much for any help
This is actually an error propagation problem. If you follow the formulas in the link I gave you, you can propgate the error such that your well averages (ie 5+/0.5) get their error averaged too when you average all 3 wells.
Thanks, will look into it.