Though the Bland Altman analysis may seem to be useless under the circumstances given, the answer is: 'depends ...':

If instrument B is accurate over the whole measurement range, instrument A can be calibrated in one point, 2 points or multi-point. But there is still a lot of points where it has not been calibrated with respect to B - leaving quite some 'space' for differences.

If I understand your question (which is not as clear as one would wish - giving too little details), you are measuring some chemical or physical property. Maybe alike pH-measurements. Thus one question come into my mind: would it be possible to make a 'reference sample' to calibrate instrument A and 'check' instrument B ?

Regards

Well Kaushik,

I do agree with what Dreher is saying.

1) What you are comparing at present is the "SAMPLES" using both instrument A (Standard Reference Calibrated) and Instrument B (as developed by you)

2) As stated by Dreher, I too agree to introduce the "REFERENCE SAMPLE" to be tested using both the instruments A and B

3) But, going a step ahead, what I would suggest is to make or test your "SAMPLES" (those 13 samples) and this "REFERENCE SAMPLE" to a third party well-calibrated instrument such as Calibration Laboratory to validate the results.

4) Kindly keep in mind that no matter what algorithm or method you use to estimate the error, you sample results remain the same.

I hope you find it useful.

Best wishes!

-Prasanna Waichal

Hi Kaushik,

In my opinion, Bland--Altman plot is appropriate in your case. It will allow you to estimate the agreement between instrument A and B, regardless of your calibration. For instance, if instrument A always overestimates the result of instrument B "between" calibration points, this will show up in a Bland-Altman plot.

The percentage error you calculated between A and B is going to give you instrument A's accuracy. The last step is determining your instrument's repeatability: you could test-retest some samples to determine it.

Good luck!

Claudio

Dear Kaushik,

Bland -Altman plot is giving the sense of agreement vs Co-relation. You have developed a instrument and you have calibrated the same with a better standard. So the reading of your developed instrument will definitely follow the standard one i.e both of them will be co-related within the acceptable error.

As per definition, it says "Bland and Altman make the point that any two methods that are designed to measure the same parameter (or property) should have good correlation when a set of samples are chosen such that the property to be determined varies considerably. A high correlation for any two methods designed to measure the same property could thus in itself just be a sign that one has chosen a widespread sample. A high correlation does not necessarily imply that there is good agreement between the two methods.

So if you compare your developed instrument with other equipment to measure same parameters (But not the calibrating equipment) then Bland -Altman plot is giving the sense of agreement vs Co-relation. So your thought process is right. So in this case this discussion will not be useful in sense.  So my suggestion use other equipment to test the result.

Best wishes

Prof. J. K. Roy

Thank you everyone for your responses.

@ Dr Dreher - Thank you for your response. I am measuring the glycated hemoglobin HbA1c value from a blood sample using our own laser based lateral flow immunoassay analyzer(dry chemistry technique) and the reference is Bio-Rad HPLC(high performance liquid chromatography) a wet chemistry technique. The two techniques are entirely different. reference sample" is not possible in my case as I am using blood as a sample and this cannot be synthesized at least not by me. I am working in developing the reader instrument.

@Dr. Roy- Thanks for your response and explanation. I found out from literature - Krouwer 2008, that a modified version of Bland Altman plot may be reported in this situation where the reference instrument readings lie along x-axis and the difference between the two lie along y-axis.

Hello Kaushik,

In that case if you can't have a reference sample, why don't you measure your HbA1c value using another technique established pathological technique. That way you can compare your own results with two different standard measurement techniques.

Best wishes!