Robustness

[syx]

Dear members,
I need to know how to perform graphical/statistical analysis of effects in robustness test. Graphically, the effects importance can be verified from a normal or a half-normal probability plot. I do not know what and how...
How to estimate error from dummy effects or from two-factor interaction effects?
I read from some literature, e.g. Robustness Test and Guidance for Robustness/Ruggedness Tests in Method Validation, but I still could not understand...


regards,
siswanto

[Mattias]

Usually you don't need to make it as complicated as stated in the articles.

I typically use resolution as my "response", and then I vary the relevant parameters (pH, temp, batch of column etc) according to a factorial design. The random error can be estimated by doing several injections without any changes.

Don't overdo the parameter changes, just be slightly outside the normal variations. If the resolution is OK in all experiments, then I don't perform any mathematical evaluation at all. The specificity is proven to be robust!

If there is a problem I go for the main effects. The results can be used to tighten parameter limits of the analytical method. I have so far not seen any two-factor interactions that have been relevant for the methods. (remember also that the authorities must understand what you have done!)