Any number you want: the impact of data cleaning on internal validity

Concerns about the internal validity of reconviction studies tend to focus on factors
such as initial comparability of groups. Often overlooked is the impact that data
preparation can have. Data preparation refers to the decisions taken by researchers
regarding which offenders to retain in the sample for analysis. Using data relating to
a sample of offenders in two police forces, it is shown that these decisions, even when
applied equally to both groups, can impact differentially on reconviction rates,
weakening a study s internal validity. Implications of the findings are considered and
recommendations made to improve the transparency of the process.