SAGE Journal Articles
This article is an alternative to Newman (2014) below, but in an education research context. There is a good review of the sources and consequences of missing data as well as a largely non-technical review of the main mechanism for handling missing data.
This recent article gets a bit technical in places (which can be skipped), but gives a good overview of the range of approaches to different kinds of missing data (see Table 4 in particular) and suggests five practical guidelines.
A persistent problem in the design of bipolar attitude questions like Likert items is whether or not to include a middle response alternative such as ‘Neither agree nor disagree’. These authors used follow-up probes administered to respondents who initially selected this alternative to see whether they selected it in order to indicate opinion neutrality or to indicate that they do not have an opinion on the subject. They found that the vast majority of such responses turned out to be ‘face-saving don’t knows’ even if a separate ‘Don’t know’ option was available. This article is interesting both for its substantive findings and for the design of the research, which follows ‘classical’ lines.