Chemical Forum

 Retrieve Password
 Register Now
!show!: 679|Reply: 2

Agilent Chemstation Calibration Curve [Copy URL]

Rank: 8Rank: 8

Post time 2012-12-22 08:17:53 |Show all posts

I am working with Agilent Chemstation and am trying to analyze for 3 compounds. I have created an 8 point calibration curve and have a good linear regression for each compound. My calibration curve ranges from 0.05 ug/ml to 10.0 ug/ml. When I go to reprocess my calibration curve for 0.05 and 1.0 ug/ml I receive amounts such as 0.00 for the first compound, 2.3e-3 for the second and 3.4e-1 for the third compound at standard 0.05 and for 1.0 ug/ml I receive 1.1e-3 for the first compound 2.4e-2 for the second and 3.4e-1. At first I thought maybe I should recalibrate or make a new curve but when I do the first two cal points (0.05 and 1.0) and then go back to the first cal-point 0.05 I get the results (5.5e-2 5.4e-2) that I am looking for. Another frustraiting issue is that samples that are in the middle of my cal-range (1.0 ug/ml) come as expected. Also when I go to calculate other samples at 0.10 concentration the area is the same as my standard in the calibration curve (44.372) but the concentration amounts are not the same. Any clues as to how I might be able to fix this?

Use magic Report

Rank: 8Rank: 8

Post time 2012-12-22 08:17:54 |Show all posts

If you use ordinary least squares for your regression over a wide range, errors (amounting to garbage) at the low end are common. Either a weighted least squares or a transform to log-log can help, but you really need to look at the distribution of residuals to get more information.

Use magic Report

Rank: 8Rank: 8

Post time 2012-12-22 08:17:56 |Show all posts

You can also get a very nice R-squared value and still have a line that misses the low-level calibration points by a considerable distance in relative terms, or even crosses the y-axis quite a way from zero, which means you can get negative answers. Tom's right, look at residuals if possible. At the very least, zoom in on the low end of your calibration curve and look where the line actually is, relative to the points.

Use magic Report

You have to log in before you can reply Login | Register Now

Archiver|Chemical Forum

GMT+8, 2017-2-22 20:56 , Processed in 0.063842 second(s), 10 queries .

Powered by chemical forum