Rejecting Data

[kristind858]

Hello !
Can anyone advise me on what to do about a rejecting an injection? Would I use the Chauvenet Criterion ? I ran a sample and prepared two sample preps : 1A and 1B. Each sample prep is split into three separate vials and are bracketed by a standard. The system suitability passed at <2% for the entire run (5 injections of standard prior to sample injection and 2 injections of standard after the last sample run). For my first three injections of 1A, the HPLC integrated : 18111; 54944; and 54948 AU. For the second sample prep 1B, the HPLC integrated around 55000 for all three injections. How would I reject the 18111 AU injection? It seems as though an air bubble may have got onto the column. Thanks for your help !!!

[tom jupille]

VERY crudely speaking, you calculate the mean and standard deviation for all six injections. If you have one that is more than three standard deviations from the mean, then there is better than 99% probability that it is an outlier.

[DR]

Assuming you're working in a regulated environment...

Once you've consulted your standard operating procedures on the subject of "out of specification results and investigations thereof" (OOS), you should have a good idea as to whether a result can be pitched based on a Rho test or not. If not, there should be mention of reinjecting, doing more preps, investigating various factors that could have resulted in the low value etc. that all must be addressed.

[kristind858]

Thank you DR and Tom !!