Explanation of splines?

amzipkin

Hi folks,
Just a (quick?) question this time around. I am looking for documentation or an explanation of how the various spline options in Iolite work. In the past I have used the default spline_autosmooth for trace element data reduction without issue. However, now I am doing 87Sr/86Sr isotope ratio data reduction using a bracketing standard for correction. Iolite is consistently overcorrecting my samples (actually secondary standards) and producing final results that are too radiogenic. For solution MC-ICP-MS 87Sr/86Sr data, I would just build a linear model (or models) of how my bracketing standard is behaving and apply a final correction or corrections in Excel. For the laser ablation Sr work, I am trying to keep everything in Iolite since in principal it can save me the trouble of doing the final corrections manually. I think the spline fit to my standards may be part of the problem but I do not have an adequate understanding of what I Iolite is actually doing with the various options available.

Joe

amzipkin

The smoothing spline code is based on:

Hutchinson, M. F., & de Hoog, F. R. (1985). Smoothing noisy data with spline functions. Numerische Mathematik, 47(1), 99-106.

Hutchinson, M. F. (1986). Algorithm 642: A fast procedure for calculating minimum cross-validation cubic smoothing splines. ACM Transactions on Mathematical Software (TOMS), 12(2), 150-153.

The smoothing parameter is determined automatically when "AutoSmooth" is selected. See those references for the details on that. If a specific level of smoothing is selected you are effectively selecting the "p" parameter in the above references. Note that if you do not have enough selections in your reference material group Spline_AutoSmooth will default to other options: >=6 selections will use auto, 4-5 will use medium smoothing, 3 will be step linear, 2 will be linear fit, 1 will be mean.

I can have a look at one of your sessions to see if I can spot anything causing your problem. Send an example (any raw data + logs or io4 file, as well as any necessary reference material files) to the iolite support email if you want me to.

All the best

amzipkin

Joe Thanks Joe; I will check out those references. The spline fitting may be part of the problem I was having but some more work in the lab this week suggests that the larger issue is a mismatch in voltage between my bracketing standard and secondary standards. In short, lower signal on the bracketing standard causes it to be more offset from its reference value than the secondary standards with higher signal. So, the bracketing standard correction is overdoing it and actually making things worse. That's my working hypothesis anyway. Today I tried tweaking ablation parameters to voltage match all analyzed materials and am about to process the results in Iolite. If things are still looking screwy and I can do a better job with a bracketing standard correction manually in Excel, then I will let you know so that we investigate if a spline fitting issue is the cause.