Model and Data Checking

Abstract

Even the most careful and thorough experimental practice cannot eliminate the possibility of an error appearing in the data collected from an experiment. It may arise because of a simple mistake in recording or transcribing a result, be due to a fault in a recording device, or perhaps result from an undetected fault in an experimental unit or an unknown affliction suffered by a subject used in the experiment. In many circumstances, it is possible to detect serious errors in the data. In other circumstances, it can be determined in advance that the nature of the experimental design will make detection either impossible or difficult.