Chemical Forums

Chemistry Forums for Students => Analytical Chemistry Forum => Topic started by: furanosa2000 on January 24, 2013, 02:17:42 AM

Title: Re: Use of origin/zero in calibration curve
Post by: furanosa2000 on January 24, 2013, 02:17:42 AM
Hi all,

I am still lost here.. I hope everyone still in the mood to discuss this. I am not an analytical chemist, but using it all the time..

consider this:
I uses vis spectrophotometry to determine protein concentration. I then develop a series of known concentration of protein, e.g. 0.0, 0.1, 0.2, 0.3, 0.4 and 0.5 mg/mL. I then add reagents, and finally measure the Absorbance.  Before reading all the standards, I calibrate (set to 0) the instrument using 0 mg/mL standard, which is the blank. and read all the rest against it. for example I have the following data: (0.0, 0.000) (0.1, 0.122) (0.2, 0.254) (0.3, 0.382) (0.4, 0.512) (0.5, 0.624)

So, my question is, as the instrument has been "told" that the reagent mixture without analyte result in 0.000 absorbance, should we input (0.0, 0,000) as one of the point when we are making the standard curve??

Its not that because I have samples with Absorbance less then 0.122, but because the instrument has been set that absorbance 0 is all the reagent mixture without the analyte. If we don't include the (0.0, 0.000) point for the linear regression calculation, I tought the absorbance of the other standard should be read without setting the instrument to zero (calibrate the instrument).

I hope everyone can get what I am trying to say.. Sorry for the bad English..
Title: Re: Re: Use of origin/zero in calibration curve
Post by: Borek on January 24, 2013, 04:08:12 AM
So, my question is, as the instrument has been "told" that the reagent mixture without analyte result in 0.000 absorbance, should we input (0.0, 0,000) as one of the point when we are making the standard curve??

Yes, your zero calibration took care of the difference between the real zero and the shown zero.

Please don't hijack threads, this question was perfectly worth its new, own thread.
Title: Re: Re: Use of origin/zero in calibration curve
Post by: furanosa2000 on January 24, 2013, 11:16:19 PM
Dear Borec,

I am sorry if I spoiled the thread. I am still confuse with the diffrence of what I done and the zero calibration curve.. so I will post my question in a new thread for a new discussion.
Title: Zero origin vs blank in calibration curve
Post by: furanosa2000 on January 24, 2013, 11:28:04 PM
Hi all,

I am a little bit lost about use of origin/ zero in calibration and using a blank solution for making a calibration curve. I am not an analytical chemist, but using it all the time..

consider this:
I uses vis spectrophotometry to determine protein concentration. I then develop a series of known concentration of protein, e.g. 0.0, 0.1, 0.2, 0.3, 0.4 and 0.5 mg/mL. I then add reagents, and finally measure the Absorbance.  Before reading all the standards, I calibrate (set to 0) the instrument using 0 mg/mL standard, which is the blank. and read all the rest against it. for example I have the following data: (0.0, 0.000) (0.1, 0.122) (0.2, 0.254) (0.3, 0.382) (0.4, 0.512) (0.5, 0.624)

So, my question is, as the instrument has been "told" that the reagent mixture without analyte result in 0.000 absorbance, should we input (0.0, 0,000) as one of the point when we are making the standard curve??

Its not that because I have samples with Absorbance less then 0.122, but because the instrument has been set that absorbance 0 is all the reagent mixture without the analyte. If we don't include the (0.0, 0.000) point for the linear regression calculation, I tought the absorbance of the other standard should be read without setting the instrument to zero (calibrate the instrument).

What actually the different between what I am doing and Use of origin/zero in calibration curve?


I hope everyone can get what I am trying to say.. Sorry for the bad English..

PS: To Borec: Sorry to come up again with this topic, I just want a more clear explanations
Title: Re: Zero origin vs blank in calibration curve
Post by: Hunter2 on January 25, 2013, 01:14:13 AM
In an UV/VIS measurement you want to get the concentration only of your substance, in your case the protein. You set the instrument to zero with a blank solution to eliminate all the interferer substances. In your curve I would only put in the points measured with your samples. The given curve should intersect in the 0,0 point if worked accurate. I would not put in 0,0 from the beginning, because 0 means no substances where you looking for.
Title: Re: Zero origin vs blank in calibration curve
Post by: Borek on January 25, 2013, 03:12:40 AM
If instrument is automatically subtracting the blank for you, you should include 0,0 point in the calibration curve, as the instrument is already forcing it - otherwise there is no way to use the blank on the calibration curve and the measurement is lost.

However, when there is no automatic forcing of zero, you should not add 0,0 point to the curve, as blank absorbance doesn't have to be zero.
Title: Re: Zero origin vs blank in calibration curve
Post by: furanosa2000 on January 25, 2013, 07:30:56 AM
Thanks for the answer.

@Hunter: it will never intersect with 0,0 point. It is a very rare case that you can find something like that. The linear regression obtained, will always have intercept, which mean the straight line will not intersect on 0,0 eventhough you include 0,0 data for the calculation of linear regression.

@Borek: So, actually if I make a blank and not forcing the instrument to zero the reading, I should also include the blank reading? for example, if the blank read 0.052, so I have to input (0, 0.052) when I make the linear regression? just to make it clear.

Do you have any references that explain this? because sometimes I am asked by the other and they insists that even though you force the instrument to zero by blank, the 0,0 point is not to be included to the linear regression calculation.
Title: Re: Zero origin vs blank in calibration curve
Post by: Borek on January 25, 2013, 07:54:45 AM
@Borek: So, actually if I make a blank and not forcing the instrument to zero the reading, I should also include the blank reading? for example, if the blank read 0.052, so I have to input (0, 0.052) when I make the linear regression? just to make it clear.

Yes.

Quote
Do you have any references that explain this? because sometimes I am asked by the other and they insists that even though you force the instrument to zero by blank, the 0,0 point is not to be included to the linear regression calculation.

No, I don't have any reference. I suppose people are applying blindly the idea about not using 0,0 point, without giving it a second thought.

Mathematically the scenario in which you use all points without applying blank correction and the scenario in which you use all points (including 0,0) and apply blank correction are equivalent and will give the same results. Using all points without correction and using correction but deleting the 0,0 point from regression will give different results.
Title: Re: Zero origin vs blank in calibration curve
Post by: furanosa2000 on January 25, 2013, 08:18:57 AM
Thanks again for the expanded answer..

Yes, you are right Borek.. we will get the same answer if we zero the instrument and include (0,0) for the linear regression, and the same answer will come up if we don't zero the instrument and include the absorbance of 0 sample to the calculation.

If anyone found a reference that explain this, could you please let me know?

Cheers
Title: Re: Zero origin vs blank in calibration curve
Post by: Borek on January 25, 2013, 08:39:28 AM
Yes, you are right Borek.. we will get the same answer if we zero the instrument and include (0,0) for the linear regression, and the same answer will come up if we don't zero the instrument and include the absorbance of 0 sample to the calculation.

If anyone found a reference that explain this, could you please let me know?

This is a simple mathematical exercise, I doubt you will find a reference for that, just like you will not find the reference for the fact 2×2=4 ;)

Linear regression is just a minimalization of a sum [itex]\Sigma_i(y_i-a\times x_i-b)^2[/itex] with regards to a and b. Instrument blank correction means b=b'+c, it is just shifting of the axis. Shouldn't be hard to show a,b and a',b+c must be identical.
Title: Re: Zero origin vs blank in calibration curve
Post by: furanosa2000 on January 25, 2013, 08:45:54 AM
Yes, its true... but the decision of whether or not to include the (0,0) point for the calculation of linear regression while we force zero the equipment is common in my place. it will change the regression linear equation if you include the zero or not, and it will also change the final calculation result. That the reference that I looking for actually.
Title: Re: Zero origin vs blank in calibration curve
Post by: Arkcon on January 25, 2013, 08:47:51 AM
I had heard that some EPA regulations require adding the 0,0 into the linear fit, even if a blank is run.  If you look through EPA or other ecological regulations, you may find a reference to this procedure.  I don't expect you'll find an explanation for this procedure, because government regulatory bodies don't always justify their positions.  So, like Borek: said, sometimes people don't reference everything, given that there's no practical need.
Title: Re: Zero origin vs blank in calibration curve
Post by: curiouscat on January 25, 2013, 08:55:44 AM
If your regression is indeed robust adding or not adding that (0,0) point ought not to change results much? Right?
If it does, arguing about which procedure is right seems difficult.
Title: Re: Use of origin/zero in calibration curve
Post by: furanosa2000 on January 25, 2013, 09:21:25 AM
Its not necesarry not to change the result so much.. for example (I like using example to make things clear) I have these data (conc [mg/mL], Abs): (0,0) (0.1, 0.122) (0.2, 0.274) (0.3, 0.456) (0.4, 0.628) (0.5, 0.749)
If we include the (0,0) data in the calculation, we will get the linear regression: y = 1.5557x - 0.0174 [1]
While if we don't include the (0,0) the linear regression would be: y= 1.6080x - 0.0366 [2]

So if  we have a sample with absorbace of 0.530, we will get the concentration of 0.3295 mg/mL using [1] while if we using [2] we will get 0.3068 mg/mL. So thats about 7% difference... Do, analytical chemist accept this difference?
Title: Re: Use of origin/zero in calibration curve
Post by: curiouscat on January 25, 2013, 09:22:53 AM
Actually, shouldn't the best way be to use blank corrected values, discard (0,0) and then use a regression y=ax instead of y=ax+b?

Why is that not an option?
Title: Re: Use of origin/zero in calibration curve
Post by: furanosa2000 on January 25, 2013, 09:30:30 AM
@curiouscat: thats should be the forcing to zero things.. but how to decide that the "b" can be omitted? if there is a "b" value it tells you that the instument is giving you signal, eventhough in the absence of the analyte.. If we omit it, thats mean that we neglect the signal that actually exist.. am I wrong here?
Title: Re: Use of origin/zero in calibration curve
Post by: curiouscat on January 25, 2013, 09:38:38 AM
@curiouscat: thats should be the forcing to zero things.. but how to decide that the "b" can be omitted? if there is a "b" value it tells you that the instument is giving you signal, eventhough in the absence of the analyte.. If we omit it, thats mean that we neglect the signal that actually exist.. am I wrong here?

The unsaid assumption behind subtracting blanks is that we have a strong prior belief that "Instrument ought to give no signal in absence of concentration. Any signal is mere error."

If yes, then isn't "conc. = a x signal" a better model than "conc. = a x signal + b" (when using corrected signals)
Title: Re: Use of origin/zero in calibration curve
Post by: furanosa2000 on January 25, 2013, 09:47:49 AM
if that the case, I think we have to include the (0,0) point when calculating the linear regression.. discarding the (0,0) will spoil your assumption that "Instrument ought to give no signal in absence of concentration. Any signal is mere error." because when you have this assumption, then (0,0) means no signal when no analyte.. But, if I not mistaken, to discard the "b" value, we have to have some statistical test before we decide it.. anyone familiar with that?
Title: Re: Use of origin/zero in calibration curve
Post by: Borek on January 25, 2013, 12:12:11 PM
The unsaid assumption behind subtracting blanks is that we have a strong prior belief that "Instrument ought to give no signal in absence of concentration. Any signal is mere error."

No, the assumption is that any signal in the absence of a substance is a background that will not change when the substance is present.

Such a signal can be unavoidable in some systems (say you are measuring presence of a brown contaminant in a slightly yellow substance), and to some extent it is always present - there is no such thing as perfectly transparent cuvette and perfectly transparent solvent (even if we often assume they are perfectly transparent, and we are usually right with a very high accuracy).
Title: Re: Use of origin/zero in calibration curve
Post by: curiouscat on January 25, 2013, 12:36:26 PM
The unsaid assumption behind subtracting blanks is that we have a strong prior belief that "Instrument ought to give no signal in absence of concentration. Any signal is mere error."

No, the assumption is that any signal in the absence of a substance is a background that will not change when the substance is present.

Such a signal can be unavoidable in some systems (say you are measuring presence of a brown contaminant in a slightly yellow substance), and to some extent it is always present - there is no such thing as perfectly transparent cuvette and perfectly transparent solvent (even if we often assume they are perfectly transparent, and we are usually right with a very high accuracy).

Yes, you said it better.Even so, isn't that a stronger case to use y=ax in the regression?
Title: Re: Use of origin/zero in calibration curve
Post by: JGK on January 25, 2013, 02:02:13 PM
Actually, shouldn't the best way be to use blank corrected values, discard (0,0) and then use a regression y=ax instead of y=ax+b?

Why is that not an option?

Have you ever tried to blank correct HPLC chromatograms?
Title: Re: Use of origin/zero in calibration curve
Post by: curiouscat on January 25, 2013, 02:06:49 PM
Actually, shouldn't the best way be to use blank corrected values, discard (0,0) and then use a regression y=ax instead of y=ax+b?

Why is that not an option?

Have you ever tried to blank correct HPLC chromatograms?


Nope.