Missing values are empty fields in data, and how they are handled directly affects the outcome of the analysis. It is essential first to establish the reason for the missingness: a value that was not recorded by chance calls for a different solution than a value that is systematically missing for a particular group, and different again is a situation where an empty value actually means something, for example zero consumption. Replacing an empty field with zero is the most common and most damaging shortcut, because it distorts both averages and totals. The options are excluding incomplete records, filling in a typical value, or calculating it from another source, with every intervention documented. In reports it is worth stating how many records were incomplete, so that the reader can judge the reliability of the figure.
See also: Data quality in analytics, Validation of collected data, Data normalisation.
