Metabolomics Society Forum

Software => MS-DIAL => Topic started by: spapaz on May 13, 2020, 05:21:48 AM

Title: coefficient of variation
Post by: spapaz on May 13, 2020, 05:21:48 AM
Hi Hiroshi,

regarding the coefficient of variation (CV%) shown by the MS-DIAL normalization report.
I have been manually calculating  the QCs across samples and QCs, for some internal standards,
but I do not manage to reproduce exactly the same values.

in MS-DIAL, are the CV% calculated based on the log transformed values? (as the report shows the log scale)

I tried both with and without log transforming. When I use the log values, I use the following formula:
= SQRT (10^(LN(10)*STDEV^2)-1)
as discussed here:  10.15406/mojpb.2017.06.00200

I attach the CV% calculated on the log values (for QCs and Samples), and the corresponding ones from MS-DIAL (in red)
(Note: they are not expressed as %, but as fraction, i.e. CV 0.15 = 15%)

Thank you!


Title: Re: coefficient of variation
Post by: Hiroshi Tsugawa on May 13, 2020, 07:43:25 PM
Thanks Stefano,

the CV value is simply calculated by the original, i.e. non-transformed values although the Y-scale is finally visualized by log-transformed value. I will prepare both CV values into the figure at next update.