Handling Missing Data in Survey Research
Almost every survey has some missing answers. How you deal with them can affect your sample size, your results and how examiners judge your analysis. In this blog, we'll look at why data goes missing, the main types of missingness, and the common ways to handle it.
Why Data Goes Missing:
- Respondents skip questions they find sensitive, confusing or irrelevant.
- People drop out partway through long questionnaires.
- Technical problems or data entry errors can also create gaps.
Understand the Pattern First:
- Start by checking how much data is missing for each variable and each respondent.
- Look for patterns, for example whether missing answers cluster on particular questions or among particular groups.
- A small amount of scattered missing data is usually less worrying than a large amount concentrated in one variable.
The Three Types of Missingness:
- Missing completely at random (MCAR): whether a value is missing has nothing to do with any data, observed or unobserved.
- Missing at random (MAR): missingness is related to other observed variables, for example older respondents skipping an item more often.
- Missing not at random (MNAR): missingness is related to the missing value itself, for example people with high incomes declining to report income.
- Little's MCAR test is often used to assess the first assumption, where your software supports it.
Common Ways to Handle It:
- Listwise deletion: excluding any case with missing data from an analysis. It is simple but can reduce your sample considerably and may bias results unless data are MCAR.
- Pairwise deletion: using all available data for each analysis. It keeps more data but can make results harder to compare across analyses.
- Mean substitution: replacing missing values with the average. It is simple but reduces variability and is generally discouraged.
- Multiple imputation: creating several plausible datasets and combining the results. It is widely recommended when data are MAR, and SPSS offers it in some editions.
Prevent and Report:
- Reduce missing data at the design stage with clear questions, a sensible length and, for online surveys, gentle reminders for unanswered items.
- In your thesis, report how much data was missing, the pattern you found, the method you used and why.
Missing data is normal, but ignoring it is not. Examine how much is missing and why, choose a method that suits the pattern, and report your decisions clearly. Transparent handling of missing data will strengthen confidence in your results.