I've been doing a little light reading in "Practical HPLC Method Development" and found a reference on page 656 (2nd ed.) that states a linear std curve intercept should be through zero. I hadn't given this detail much thought. Is the intercept forced through zero, or should the y-intercept be reasonably close to zero?

I'll sign off anonymous in case this is too basic!

Just Wondering

in our lab there is a lot of discusion at this moment about intercepts. until now we allways calibrated with y=ax+b. this always gives a better fit then y=ax. but if there is no interferring peak in you blank gromatogram, b should be 0. you can statistical test if you b is significant different from 0 and if not calibrate with y=ax. somehow this "feals" unnatural but statistics learns this...
if your calibration is right so your sample-concentration is in the middle off your calibrationline then the calculated concentration will be the same ( or better not significant different). but we work in forensic toxicology where we even try to quantitate the smallest amounts. then there is a significant difference. a y=ax-callibration should be better. this is also what the external evaluation-samples with low concentration tells us....
hope this helps.

If the linearity data is being used to justify a single-point calibration, the regression line must pass through the origin. That is, make sure the 95% confidence interval for the y-intercept includes zero. If not, you may simply perform a linear calibration.