Linear regression

[sllimbri]

I am starting to use linear regression as my curve quantitation method for 8270 runs. I am confused if I should or should not force thru zero. Can anyone shed some light on this? We were getting negative quant values or a very low response would give a value > my PQL and look like a valid hit. This happened when I did not force thru zero.

Thanks.

[bartjoosen]

Inject 5 different concentrations (eg 70, 85, 100, 115 and 130% of your current concentration), make, a regression line, calculate confidence interval of the intercept. If 0 is in the interval, calibration thru zero is ok.

Bart

[Peter Apps]

If you are getting negative results you must be off the bottom end of the calibration, which is poor practise in itself. If forcing and not forcing the calibration through zero changes the result, then the line does not go through zero and you should not force it to. The zero - zero intercept is not data unless you want to include the point from a blank run.

Peter

[sllimbri]

bartjoosen,

what is a confidence interval, and how do I calculate it? We are using HP chemstation. Does it calculate it for me?

[Steve Reimer]

Since you are using 8270, start with page 42 or so in 8000C (http://www.epa.gov/epaoswer/hazwaste/te ... 00c_v3.pdf) and include a check on refit as described in section 11.5.5.2 (page 50)

This version of 8000 includes more discussion on acceptable procedures than older versions.

[bartjoosen]

A 95% confidence interval defines a range where the intercept is most likely (95%) to be inbetween.

If you do some search on the internet, you can find how to calculate this. It is possible with Excel, don't know if it's possible with Chemstation.


Bart