A Comparison of Item Non-Response in Web and Pen-and-Paper Surveys of Sexual Behaviour

Excluding respondents who fail to answer a question will reduce effective sample size. This weakens statistical power and creates problems for the creation of summative indicators, the analysis of multi-item scales and the application of multivariate techniques (Dickinson & Kirzner 1985). When item nonresponders differ systematically from responders on the variables of interest, item non-response will also lead to biased population estimates (Kupek 1999). One of the major goals of any survey design is therefore to minimise item non-response.