Hello Miss Melie, the recovery rate in the volumetric titration method is variable in the sense that if you are using alkalimetry, argentometry, Iodometry, etc, because the sensitivity depends on the normality of the titrating agent, and flag, but according to the provisions by the methods of validation of the pharmacopoeias of the U.S. and the European is + /-5%, but if what you want is to check the recovery in matrices other than drugs, and want to perform to validate some method for water analysis residual, drinking etc, this percentage of recovery may be broader. I can check the AOAC.
Dear Mrs Meile, this is relative, because is dependente of the sample and method. In according to US Pharmacopea is acceptable to pharmaceutical analysis of tablets an error of 5%, in some cases up to 10%, because manufacturing process. Glucose in whole blood, the agencies set in 5%. If you are development a new method or a new sensor and is searching for demonstrating its ability, I believe that the error can be a bit higher. However, if the method is well stablished and sample is not complex, an error lower must be acceptable. For example, moisture content in food sample must have an error about 1%.
I am working on a new method and I wonder if there are any formal regulations concerning the value of recovery %, so that i have an argument to say that may method is good.
Dear Mrs Meile, see the link: http://www.westgard.com/lesson27.htm. It describes about recovery method in clinical analysis, maybe it can help you. If you need a formal regulation, see in a Handbook about your practical. The values should be different to each area. If your area is clinical analyses, Handbook of Clinical Chemistry is the best choice. It shows the range acceptable for each analysis. You can also use the minimum square method, a plot between your data in function of the true value. If the angular coefficient is near 1, the method is not affected by proportional error. On the order hand, if linear coefficient does not start of the origin, the method presents an additive error. You must apply t-student to coefficients with 2 freedom degree. Thus, if you obtain the best response, angular and linear coefficents near 1 and 0, you can talk that your method is good. This is just a suggestion.