[jehu3]

This is easy I'm sure but I'm confused all the same.  I apologize if this is not the place to post such a question.  I'm doing a chemistry experiment and thought perhaps someone has run into a similar issue and can guide me in the right direction.

I have to plot the line of best fit on a graph to determine the slope.  I did an experiment and we're using the Beer's Lambert Law.  The line is supposed to plot this particular law. 

My question is, my first point for Absorbance (Y axis), Concentration (X axis) is of course 0,0 since it is the blank.  When I'm drawing the line of best fit, do I start this line at point 0,0?  Or, is it possible for the line to start elsewhere?

I have tried both ways and the slope differs depending on whether I start the line at the origin or if the origin is just another point on the line - inwhich case the intercept is a negative number.

Any help would be much appreciated.

[Yggdrasil]

Either way works, it just depends on the situation.  In theory, the intercept should be zero, but in practice, the measurements may exhibit systematic errors which can make the intercept nonzero.  I'd say that if the intercept is very close to zero, just make it zero.  However, if the intercept differs significantly from zero, use the non-zero intercept and make a note of the non-zero intercept being due to systematic errors in the data collection (and speculate as to what those errors may be).

[enahs]

If you are doing a Beers Lambert plot, you need to force it through zero, as it is a function of concentration and absorbance, it makes no sense having 20% absorbance at zero concentration, does it? Also by forcing the intercept (in every plot you do, not just with the blank) you standardize it, which is a good thing and makes calculations easier.

[Borek]

you need to force it through zero, as it is a function of concentration and absorbance, it makes no sense having 20% absorbance at zero concentration, does it? Also by forcing the intercept (in every plot you do, not just with the blank) you standardize it, which is a good thing and makes calculations easier.

Bad idea IMNSHO. Calibration curve is prepared to reflect the fact the the solution can already contain other substances that absorb at the same wavelength, you may have a hardware problem. It is better to avoid such things but thay can be present - and if they are consistent between measurements and the measurements reasults are easy to reproduce, callibration curve takes care of them.

Forcing otherwise perfectly linear calibration curve with high correlation coefficient to go through zero makes your results doubtfull. I am with Ygg here.

[enahs]

I agree that in principal it is not the best idea, in this case it is better then not (unless he wants to do lots of real statistic, which I personally would).

Odds are in his experiment he will also be using absorbance to determine an unknown concentration. If the 0% concentration crosses the Y axis at random points, it will be hard to determine the concentration based off of absorbance.

By forcing a point 0,0 into the linear regression curve gives you the desired effect (standard reference of 0)  and the error with this is averaged into he slope by the linear regression method.

If you run a trial, and the 0% concentration crosses at 20% absorbance, and you just assume that every point has an absorbance value of 20% less and subtract 20, well that is horrible math and statistics and will be just flat wrong. By forcing through 0,0 you are averaging your errors in with your data, which is much more reliable then the other method (as each time you are just making up random error and you are not attempting to correct for any of it).

Yes there are many better ways to handle this type of calculations and error, but I doubt the original poster wants to learn statistics.