Home Forums UPb Geochron DRS U/Pb standard practices for converting SE to SD?

Tagged: , ,

  • Creator
  • #11213

    Curious what the U/Pb community standard is for converting SE to SD for data reduced in iolite — using the typical SD=SE*SQRT(n) conversion and using number of points in the selection (e.g.), the SD values I get are unexpectedly high.. Do folks just report the SE as “2sigma”? Or did we just get some super imprecise data?

Viewing 1 replies (of 1 total)
  • Author
  • #11228

    This topic has been previously discussed in the iolite forum, but because that happened many years ago the posts are not longer available online. However, I managed to recover them from our offline backup. Below is Chad’s explanation of the values exported by iolite which I think answers your questions:

    As Jon said, if you’re getting such tiny errors in Isoplot there’s definitely a mismatch, and as he noted one option is to change the input format in Isoplot from % errors to absolute errors (the uncertainties exported from Iolite are absolute).
    The other option is to convert the absolute errors coming out of Iolite to % errors. You’re actually halfway there by multiplying by 100, but you’ve missed also making the error relative to the value (a percentage is a fraction of the total, expressed in percent). The equation for converting from absolute to percent is: Error(%) = Error(absolute) / Average * 100
    for example 200 ± 30 is an absolute error. Converted to % it would be 30 / 200 * 100 = 15%.
    (This is why your 206/238 errors are still off by a factor of roughly 10, you need to divide by the average 206/238 ratio, which is roughly 0.1).
    Regardless of which way you choose to get the errors from Iolite and Isoplot to match up, it’s important that you keep the Isoplot input as 2 sigma, as all of the uncertainties produced by Iolite are expressed at the 2 sigma level.
    And lastly – it’s quite understandable that you find the 2sd and 2se confusing. The numbers coming out of Iolite are indeed 2se. The 2se is the best estimate of the uncertainty on the average value, which becomes increasingly well constrained as more data is included.
    As Jon noted there is then the additional problem of determining how accurate the average value is – it’s all well and good to collect data for so long that the 2se becomes fantastically small, but that doesn’t necessarily mean that the average is getting any more accurate… And this is where long term reproducibility and uncertainty propagation come in, as they are efforts to determine the limit on accuracy, which is often much larger than the 2se (internal precision) of a given analysis. Our propagated uncertainties are an effort to take into account the limits on accuracy within a given session, and to inflate the quoted uncertainty to reflect this if required.
    Anyway, regardless of all that detail, the important thing to note here is that the 2 se error coming out of Iolite is our best estimate of the uncertainty on the average for that analysis, and thus the right number to feed into Isoplot.
    I hope that helps? Cheers, Chad

Viewing 1 replies (of 1 total)
  • You must be logged in to reply to this topic.