Some Guiding Principles

The estimate of method variability from (i) a method control chart or (ii) the root mean square error (RMSE) from a regression model fit to a reasonable amount of stability data (from lots stored at the recommended storage condition) will provide reasonable long-term estimates of method precision (standard deviation).

Justification: general discussion in this USP Stimulus Article

Method qualification or method validation designed experiments underestimate method precision (standard deviation) owing to the limited number of runs executed in the designed experiment.

Justification: general discussion in Schmelzer et al

Owing to the possibility of biopharmaceutical results being serially correlated, at least n=30 is required to obtain reasonable estimates of central tendency and variability

Justification: this article

Owing to its ease of interpretation, an individuals control chart - used in conjunction with Nelson run rules - is more practically effective than implementing a CUSUM (Cumulative Sum) or Exponentially Weighted Moving Average (EWMA) charting program.

Justification: none (solely Keith Bower's opinion)

If there are at least five different numeric values, it is somewhat reasonable to consider the data as continuous, thereby calculate estimates of central tendency and variability to use in control chart limits.

Justification: (tangentially) AIAG Measurement System Analysis (MSA) Manual (regarding the number of distinct categories)