Designed Experiments With Missing or Discordant Values

摘要

THOSE who have carried out designed experiments have occasionally encountered a missing value or a value that was clearly discordant. How does one carry out a proper analysis in such a situation? When a value is missing from a balanced experimental design, such as a factorial design, it is desirable to substitute some value for the missing one so that an analysis can proceed. It is possible to write a model for the unbalanced design and use a regression analysis, but this requires advanced knowledge of this subject. Otherwise, an appropriate value must be chosen to substitute for the missing one, thereby permitting a standard analysis to be carried out.