Hi, see the Guidance from Eurachem 2014 or NATA 2013, there also many other Guidance's for validation methods, I think you can by searching via Google.

You can also review USP section .

Gunawan and Bruce, thanks for your response to my question. I certainly read many papers and publications regarding this issue but I could not find the maximum acceptable values for CV and DEV% when validating a new method. Have you found that?  

No.  It is common in 'method development and validation' that variance is expressed in %RSD.  It's just another side of the same coin!

Remember.  The sample (and thus the method results) distribution is Gaussian and the 'standard deviation (SD)' as the expression is appropriate.

The distribution for the variable to be normal is not a necessity. If you take N, Mean and SD of several studies/experiments, you will land-up in handful of patterns which cannot be generalised. It is my experience that you have to decide the limits on your own logic.

However, in a paper by JK Pael, NM Paten and RL Shiyani have reported the following:

The upper fiducial limit of CV (based on average degrees of freedom) was 21.78%. This value was covered in the class 21–24% (including 23%, the yardstick). It can be seen from Table 2 that more than 50% of the experiments having 21% and more CV failed to detect difference in treatment means. Results clearly showed that when the coefficient of variation in pulse experiments in RCBD exceeds 23%, the experimental findings should not be considered for any purpose. This upper fiducial limit (23% CV) is based on truncated t-distribution and the theory of probability and is therefore, recommended for further use as a yardstick for CV of field experiments, to be conducted in RCBD on pulse crops, in Gujarat. For generating an acceptable yardstick for CV of agricultural field experiments, it is suggested that the large number of experiments (> 10,000) conducted on different crops be evaluated for this purpose.

Basically CV30 is not acceptable.

Acceptance values depend on the variation in the sample matrix and the analytical method and are relative to the specification. For example, a API assay specification between 95.0-105.0% of the label claim may have an %RSD of 5.0% or less. However, if the label claim is 98.0-102.0% then this acceptance criteria is too high and 1.0-2.0% RSD may be more appropriate. There is NO rules or guidances on setting acceptance criteria but you can review USP on chromatography of pharmaceuticals and biopharmaceuticals. This means you may have different acceptance criteria or specifications for the API assay, related substances, cleaning validation, process validation, dissolution, clinicals...

The above advice applies only to pharmaceuticals and biopharmaceuticals. This means you must prove the method is appropriate to your sample matrix. For example, it only takes 1 corn kernel to contaminate above the limit an entire truckload of corn with a mycotoxin. It's CRAZY to grind and test an entire truckload when many elevators have an adequate cleaning procedure to reduce this RISK.

Dear Mr. Abdelrahman. I agree with Mr. Ehsan Ebrahimi. In addition, the SD range is part for calculating CV value. Hence, CV is important value for detection level variation in the data population. Thanks

agree with@ Widya Putra

regards

Ehsan Ebrahimi , do you have a reference for this: " Basically CV30 is not acceptable"? If you have, can you give me, please?

Thanks!

@Elren Monteiro & Abdelrahman Zaky

The acceptable CV level is depended on the spirit of the research. For instant, in clinical chemistry research, acceptable CV is different with field experiments in agriculture. But in overall you can check: STATISTICAL PROCEDURES FOR AGRICULTURAL RESEARCH (Second Edition) by KWANCHAI A. GOMEZ or you can have a look at:

http://clinchem.aaccjnls.org/content/clinchem/35/4/630.full.pdf

Ok! Thank you so much Ehsan Ebrahimi !!

Hi everyone and thank you for the interesting discussion!

I would be happy if someone could refer me to papers which had detemined coefficient of variation cutofffs/acceptable levels, in any field. I study the medical practice variation field and use the coefficient of variation method, and therefore searching for accepatable levels. Ehsan Ebrahimi has referd an acceptable level in the clinical chemistry and agriculture fileds. I would be happy to find acceptable levels in more fields.

Thanks in advance,

Sagi

Thank you so much Ehsan Abrahimi

It is different from case to case but generally, CV value between 2% and 3% is good and acceptable.

Thanks to everyone, for your valuable contribution to this question. I would conclude that the acceptable CV differs based on the research sector. In medicine and pharmaceutical research this CV should be very low

I have an idea of choosing a suitable COV limit in my research area, civil engineering. Usually, the standard or code or guideline will have a safety factor to make the calculation more conservative. If the factor is 0.9 or 0.85, we may allow the COV to be less than 10% or 15% so the largest variation will not make the structure unsafe after the factor.

Dear all can you give some paper references about the acceptable range of percentage deviation (DEV%).

How about the acceptable value for the CV for chemical analysis, is it same as medical field?

Nattira,

It depends on the risk to the final consumer. What is the purpose of the chemical analysis? Would you take a drug that may have no Active Pharmaceutical Ingredient (thus not effective)?

depends on the case of study and available data for wet land because it black box in treatment and transformation of the micro-organsims the coefficient of variation greater than 20

Dear all, Thank you for good answers! Can I get the answers for interpretation?

Ehsan Ebrahimi,

For the "CV30 is not acceptable." categorization, is this referring to the agricultural research thresholds or the clinical chemistry one? I would suspect that the clinical chemistry one is lower. Unfortunately, I do not have access to the clinical chemistry paper.

where can I get the citation for this?

Basically CV30 is not acceptable.

Thanks

Cause of very low variation, I have seen a CV of less than 1. Shivany Condor M. Nasir Shalizi Shankarappa Hanumaiah

Comparing direct vs indirect VO2max by heart rate deflection in soccer players I reached 4%. A great estimate.

Definition of CV: The coefficient of variation (CV) is the standard deviation divided by the mean. It is expressed by percentage (CV%). CV% = SD/mean. CV30 is not acceptable.

CV should be less than 10.0 for good statistics. The following webpage will be useful for you as well:

In research, the CV should be less than 10%.

CV30 is not acceptable. please suggest reference if any?

Aronhime, S., Calcagno, C., Jajamovich, G. H., Dyvorne, H. A., Robson, P., Dieterich, D., ... & Taouli, B. (2014). DCE‐MRI of the liver: effect of linear and nonlinear conversions on hepatic perfusion quantification and reproducibility. Journal of Magnetic Resonance Imaging, 40(1), 90-98.

....by calculating the coefficients of variation (CV = SD/Mean × 100%). We defined parameter reproducibility as excellent when CV was ≤10%, good when CV was between 10–20%, acceptable when CV was between 20–30%, and poor when CV was >30%.

CV30 is not acceptable

Please references

Is these levels same for the field researches?

acceptable values for the percentage deviation (DEV%) is depends on each analyte measured and is taken from published data

acceptable values for coefficient of variance (CV) is < 5%