i always think the software validation does not apply for those software that does not perform calulation. if the software is used
to compare a reference standard and the testing part, then the reference standard need to be calibrated but not the software.
Software which used to calculate statistiscal parameter, following by analysise need to be validated. Statistic are complex, you need to vailidate the assumption make by the sofeware during calculation with your assumption.
For example: Is the software , make use of sample sigma or esimate of process sigma for calculation, what type of distriubtion they assume: t-distribution, normal distribution, What level of confident they use, 95%, 90% when they calucated the limits. When calucating process capability, are they use six sigma or traditional three sigma philosohy.
This just what i think. I do not know if i am right. Pardon me if all these are rubbish. Maybe someone can provide a better answer.