We have a large number of samples whose concentration we measure twice, averaging the two values. Typically, the coefficient of variation (cv) for each sample is < 5%, but for a few samples the cv is high. We assume that in these cases something went wrong with one or both concentration measurements. We can afford one more concentration measurement for the samples with high cv's.

My question is, how to use the three measurements to achieve the "best" estimate of the true concentration? Average all three measurements? Pick the two with the lowest cv? Or...?

Many thanks for any insights or pointers to literature.

-
just to make sure, you calculate cv for the sample only from the two measurements? –  mpiktas Mar 30 '11 at 21:53
Picking the two closest to each other and averaging them seems reasonable. But do you know what exactly goes wrong and why? If you do, this could influence how you do this. For example, if you know the measurement sometimes gets stuck at 0, then you may want to just ignore any zeros. –  SheldonCooper Mar 30 '11 at 21:55
Or you could go for the median of three (in a sense you already choose the median of two) –  Henry Mar 30 '11 at 23:52
@mpiktas: yes, we calculate cv from only the two measurements. @SheldonCooper: we usually do not know what goes wrong (but the stuck at zero idea is a good one to explore). –  Ed Hagen Mar 31 '11 at 2:24