noahmclean
commented
8 years ago Before the wrong nomenclature propagates too far... these aren't 'error ellipses', despite the nice alliteration, they are uncertainty ellipses. The distinction between 'error' and 'uncertainty' is both important and useful in metrology -- the modern science of measurement, a branch of statistics.
Short Story: An error is the difference between the measured value and the true value. For instance, check out the paper here: http://www.mantleplumes.org/WebDocuments/Renne1997.pdf . In it, 1,918 years after the 79 AD eruption of Vesuvius (in 1997), pumice from that eruption was dated to be 1,925 years old. The error here therefore is 7 years. Note that this is a special case -- we usually don't have human records for the events that we're dating, like the extinction of the dinosaurs 66 million years ago, so usually we don't actually know the error. We can just make a (statistical) estimate about what it might be. An uncertainty is the range of values (usually stated with a ± ) that could reasonably be assigned to that age, with an estimated probability that the true value lies in this range (often part of a measured or assumed probability distribution, e.g. Gaussian). In the Vesuvius paper, the uncertainty is ± 94 years, so that the date is reported as 1925 ± 94 years ago. Rather importantly, I can't find a hint in that paper whether this is a 1σ or 2σ (68% or 95%) uncertainty -- it is probably 1σ, as this is the convention for Ar-Ar dates (but not for all other isotope systems, who prefer 95% confidence intervals).
Long Story: The VIM, or international vocabulary of metrology, is the go-to glossary for definitions, though sometimes the English writing there can be a bit muddled. Go to http://jcgm.bipm.org/vim/en/ and navigate to 2.16 for the definition of 'measurement error' and to 2.26 for 'measurement uncertainty'. I think this is a useful distinction, though I don't really find the way they distinguish accuracy, precision, and trueness to be very helpful.
johnzeringue
bowring
In the latest versions of Topsoil (the v1 alphas), it is no longer possible to configure the size of the uncertainty ellipses, because that functionality has yet to be reimplemented. At the moment, all uncertainty ellipses plot as 1σ ellipses. @noahmclean has provided an excellent explanation of what this means.
November 2nd
Updated based on comment below.