If you refer to an analytical sequence where you run bracketing standards, QC samples, and pairs of replicates of regular samples I would not worry too much about a standard deviation calculated on two replicates even if you run them one after the other because 1) the samples may not be isotopically homogeneous, 2) you need more data/replicates to describe a population through statistics, 3) there may be combustion problems, and 4) memory effects may affect the first sample of the pair. I would start worrying only if what you describe affects pairs of replicates of isotopically homogeneous standard materials and/or the standard deviation does not get better after analyzing more replicates. Of course, I assume that there is no problem with the overall analytical precision and accuracy as calculated on QC samples intercalated among the regular samples. Cheers. Mihai
Dr. Mihai Lefticariu
Mass Spectrometry Facility
Southern Illinois University Carbondale
Carbondale IL 62901
From: Stable Isotope Geochemistry <[log in to unmask]> on behalf of MVT <[log in to unmask]>
Sent: Thursday, March 3, 2016 6:40 AM
To: [log in to unmask]
Subject: [ISOGEOCHEM] Calculating duplicate error for C & N samples
I was wondering if there was a general consensus within the forum as to the correct way to calculate standard error between replicate samples for carbon and nitrogen analysis. Until recently we would calculate the standard deviation of the two replicates, and run a third sample if the value was greater than 0.2 per mil.
I have recently been informed that calculating the standard deviation of two values is not actually correct and we should be calculating absolute error? Can anyone shed any light on this, I would be interested to see what other labs do when working out whether the replicate data is usable?
Many thanks in advance