# 10.1.1. Dealing with missing data

The treatment of missing data is a rather specialised statistical topic. Missing data is difficult to deal with adequately in the analysis of questionnaire data, especially if it is not "missing at random". Data which is missing at random is such that the responses that would have been given are not related to the probability of non-response. If data is missing at random, then the data that is available can be analysed and the results should still be representative of the population, provided that the selected sample was representative. If it is not missing at random, then the results of analysing the available data are likely to be badly biased. Missing data reduces the number of responses available to analyse and hence reduces the precision of any estimates made. The best approach therefore is to try to minimise the chances of data being missing, by careful questionnaire design and by choosing a survey mode which gives respondents time to complete all the questions and secures their co-operation to do so.