Assess uncertainty for each sample

[twollnik]

As PhD student and user of the data, I would like to be able to asses the uncertainty for each sample (=standard deviation > 0.1 permil) in order to identify bad measurements and wrong data points.

Wir möchten Ausreisser innerhalb einer Probe – also eine einzelne Injektion, die rausfaellt – angezeigt bekommen, um die Möglichkeit zu haben, diese Messung ignorieren zu lassen (evtl. auch über das User Interface auszuwählen).

[twollnik]

open question: What does "sample" mean in this context? (One injection, one vial, one rack?)

[thomas-muench]

Sample

Sample always refers to the snow/ice collected at a certain position in the field; e.g. the sample from the ice core "B32" at the depth 120-122 cm. For the measurement this sample is thawed and put into a vial.

Injection

Each sample (water in the sample vial) is injected some number of times (typically, 3 or 6 injections) into the measurement device during the measurement process. Each injection creates a data point in the output file, which is actually an integrated value of measuring the water from this injection over a certain amount of time. Therefore, a standard deviation (SD) value can be assigned to this data point, which is written along with the data point to the output file. This SD value is a first measure to assess the goodness of the data point for this injection.

Sample value

The isotopic value of the sample is obtained from averaging over some number of injections and an SD value for this average value can be calculated. This SD is then a measure for the overall goodness of the measured value of the sample and can tell you how stable consecutive measurements are or if there are outliers.

[twollnik]

@thomas-muench Thanks for the well-written explanation. This answers my question for the time being.