Key points at a glance
- Distinguish missing responses from questions intentionally skipped.
- Choose a method that fits the data, analysis and defensible assumptions.
- Report missingness and the actual sample size for each analysis.
What does an empty value mean?
An empty cell may mean a participant did not answer, a filter skipped the item or information was lost during import. Codes such as 99 may also represent missingness. Clarify these meanings before calculating statistics.
Check the questionnaire, codebook and raw data together. “Not applicable”, “do not know” and “prefer not to say” are not automatically interchangeable. A genuine zero must not be classified as missing. Record the mapping before recoding variables.
Investigate the amount and pattern
Count missing values by variable and identify records with many gaps. Examine whether missingness clusters in particular question blocks or observable participant groups. This can reveal technical problems or systematic differences that deserve further attention.
A small proportion of missing values does not guarantee that they are harmless. Equally, not every gap requires the same treatment. The plausible process behind missingness and the requirements of your analysis matter more than a universal percentage threshold.
Choose and justify the handling method
Analysing complete cases may seem straightforward, but it reduces the available sample and can introduce bias under inappropriate assumptions. Different sample sizes across analyses can also complicate comparisons. Automatically replacing missing observations with a mean is not a general solution.
Imputation procedures and models that accommodate missing data require justified assumptions and correct implementation. Discuss the choice with methodological support when needed. Explain the rule and its implications, not just the software menu command.
Preserve the raw data and implement cleaning transparently in a working copy or script. This allows you to check whether changing a decision later actually changes the results.
Report decisions and usable observations clearly
Original numerical example: Of 120 respondents, 108 answer a satisfaction item. A relationship involving a second variable has 101 complete pairs. Report the appropriate count for each analysis; the initial total of 120 does not describe every statistic you calculate.
State missing values, exclusion rules and any replacement procedures. For the observational study designs it covers, STROBE requests reporting of missing-data handling and missingness in relevant variables. This illustrates a useful transparency principle rather than a universal requirement for every thesis.
Final submission context
Formal details can feel like a separate writing task, but they matter just as much for a printed submission. Anything missing, misplaced or inconsistently formatted in the document will also appear in the bound copy.
Use this guide together with your cover page, table of contents, page numbers, source notes and appendices. Prepare the final PDF for printing and binding only after the complete file has been checked.
Practical check before PDF export
- Do headings, chapter structure and page numbers match?
- Are sources, figures and tables included completely?
- Are required elements such as declarations, appendices or the cover page included where required?
- Did you open and check the final PDF after exporting it?
Checklist
- Distinguish missing-value codes from genuine zeroes.
- Identify filter-related skips.
- Inspect the amount and pattern of missingness.
- Justify the chosen handling rule.
- Report the actual sample size for each analysis.
Common mistakes
- Replacing every empty field with zero.
- Using mean replacement without examining its consequences.
- Reporting the original total as the sample size for every analysis.
Print and bind your finished thesis
Once you have checked the content and final PDF, configure the printed copies to match your submission requirements.
Configure printing and bindingRelated content
Explore related guidance on thesis structure, sources, formatting and final submission.
Frequently asked questions
What percentage of missing data is problematic?
No threshold works for every analysis. Even a small amount can matter when missingness is systematic. Assess its amount, pattern, plausible causes and the assumptions of your method together.
Can I delete incomplete records?
This may be defensible for some analyses, but it is not a universal default. Justify the rule, examine the remaining sample and discuss possible bias.
Where should I describe missing-data handling?
Explain the rule in the methods, report missing observations and usable sample sizes with the results, and discuss relevant consequences for interpretation in the limitations.