Missing data occurs when there is no recorded value for a particular variable in an observation. Missing data may occur if a variable was not measured or reported. For example, in survey research participants have the option to skip questions, which can result in missing data.
Where possible, one should try to avoid missing data. When missing data occurs, you need to establish a standard method to identify it as such.
- Depending on the analysis software used, one common approach is to code to identify missing data; using -999 is a common convention. This is helpful especially if there is the possibility that data may be exported to another format – when exporting data, some programs automatically fill missing values (e.g. in these instances you need to verify that blank cells were not inadvertently filled). This can be avoided by coding missing data in your original dataset.
- Indicate the code(s) for missing data in the metadata/codebook.
- Do not use zeroes to represent missing data – this can lead to misinterpretation.