Rejection of data

Data for individual calibration, QC and study samples may be rejected based on assignable cause. Rejection of data due to assignable cause includes documented sample processing errors or when there was a suspected hardware failure (e.g. autoinjector, chromatography, interface or mass spectrometer problem). Assuming that criteria have been established prior to the analytical run, data may also be rejected for other observed conditions, such as when unusual internal standard response is observed, e.g. samples with double the expected internal standard area or samples with no internal standard 

Rejection of runs for other reasons must be scientifically justified all rejected data must be documented together with the reason for the rejection. 

---

These correlations are of generally poorer precision than those for reactivity and F-nmr shift data, and in the case of the ortho data set they required rejection of data for the substituents CHjCO, NO2, and CO2R (which give large deviations if included). Ihe latter deviations may be associated with ortho chelation effects. In any case, the Kj and X patterns, as noted above, appear essentially as expected. 

Rejection of data where the Arrhenius plot is not linear will make large differences to the predictions. Elimination of higher temperature data may be justifiable on the grounds of different reactions taking place. Data can of course also be rejected for WLF analysis. 

Multivariate IQC. Multivariate methods in IQC are still the subject of research and cannot be regarded as sufficiently established for inclusion in the guidelines. The current document regards multianalyte data as requiring a series of univariate IQC tests. Caution is necessary in the interpretation of this type of data to avoid inappropriately frequent rejection of data. 

A relatively complex area to address is procedural SOPs for chemical analyses in metabolism studies. One approach is to address the major operations that are common to the studies, for example characterization of metabolites in soil. The SOP can describe the general process, options available in the process and requirements for acceptance or rejection of data. Study-specific procedures that complement the SOPs can be outlined in detail and retained as part of the study records. These study-specific procedures can be prepared in the form of a work sheet and used for entering original documentation, such as the person who performed the procedure and the date it was performed. 

Validation runs may not be re-injected simply on the basis of failure to meet the method targets for precision and accuracy. Data for an entire validation run may be rejected only if there is an assignable cause. Rejection of data due to assignable cause includes documented sample processing errors or for other documented reasons, e.g., the data were lost or corrupted during acquisition or data processing or there was a documented hardware failure. Complete documentation of the failed run and the reasons for the failure should be maintained in the study file. 

Data quality objectives (DQOs) are a critical component of any project requiring the extensive collection of raw and analytical data. This was especially so with this project due to the extensive field sampling campaign (including the collection of a large number of critical field data parameters) and the importance to the health and safety of a large number of employees involved in the project sponsor s gas operations. To support this effort a detailed Quality Assurance Project Plan (QAPP) was developed to inform and guide the project team so that data of the required quality would be produced and a clear protocol for the acceptance/rejection of data was provided. 

Rejection of data simply because it does not agree with other data must be handled carefully. Some people will choose to reject one or two data points that appear statistically to be erroneous others feel that they should be included, but enough additional data taken so that the wild values do not contaminate the average significantly. It is agreed that you reject data only once - do not reject some data, then look at the new statistics and reject more data, and so on. 

---