Error Calculations in Probe for EPMA

[cvidito]

Yesterday I had a former user of the Rutgers Jeol JXA-8200 e-mail me a question regarding the errors output from Plot! in Probe for EPMA. She asked if the % errors for the elements can be used to calculate error bars for the wt% or elemental concentrations output from Plot!. I didn't think this was a problem until I saw in the software manual that the calculation for the analytical error (1 sigma) is quite different from the calculation for the error bars (also in Plot!). Can the analytical % error be used to calculate error bars or will the error bars used in Plot! have to be output from the program somehow? Thanks.

Chris Vidito

[John Donovan]

Hi Chris,
Great question.

I would use the Output | Save User Specified Format Output menu and select the Output Analytical Errors in Relative Percents checkbox. Most plotting programs (e.g., Grapher) will allow one to select error bars from a data column in relative percent.

The reason to use the analytical errors calculation is because the program utilizes both the unknown and standard peak and background intensities for properly propagated errors due to counting statistics. And this can be calculated for a single datum.

Of course there are probably additional random errors but those are more difficult to quantify. But assuming that one has acquired replicate data and that the composition in question is homogeneous, one can simply use the standard deviations (also available from the Save User Specified Format Output menu) for the purposes of errors bars. 

This latter method has the advantage of including all sources of imprecision in the error bars.

But make sure you have the most recent version of the software (10.1.x)  which can be downloaded here:

http://www.probesoftware.com/Update.htm

john

[cvidito]

Thanks John. I will do that in the future. But in this case she doesn't have Probe for EPMA (or the MDB files) and I am away from Rutgers for several weeks. Is there any reason why she couldn't use those analytical errors to calculate her error bars? I just have to make sure for her sake.

Chris

[John Donovan]

Hi Chris,
I'm not quite following you.

Which analytical errors are you asking about?  The ones displayed in the Plot! window?  Did you print a hard copy of that plot for her or something and now you want her to measure them off a piece of paper?

That is possible and would provide a very crude 1 sigma error, but it doesn't include the background and standard intensity variances.  If it is major elements of course it's not such a big deal, but for minors and traces it becomes very important.

It would be easier and better for all if you just sent her the CalcZAF and ProbeForEPMA msi files and her MDB file (and your Standard.mdb file of course). You (or she) can download the msi files from our Update web page, but you'll have to send her the run data and standard.mdb.

In the future you should always give everyone a copy of our software for re-processing purposes.  It is part of our license agreement that everyone who uses your instrument can have a copy of the software for data re-processing.

[cvidito]

John,
    Sorry about that. The analytical errors I am referring to are those which can be output from Plot!.
Chris

[John Donovan]

OK. So if she can output from Plot!, she can output from the Output menu and the latter method is more rigorous.
john

[Probeman]

Here (attached below) is a short "white paper" on what happens to "analytical error" as the interference and blank corrections are applied to a quantitative analysis: