Missing values appear when some entries in a dataset are left blank, stored as NaN or None, or marked with special strings such as "Unknown". If they are not handled properly they can:
- reduce the accuracy of a model,
- create bias in the results,
- shrink the usable sample size,
- break algorithms that require complete data.
Why handle them properly?
- Keeps models accurate — less guessing, more real information.
- Prevents bias — avoids favoring the rows that happen to be complete.
- Maintains sample size — you keep as much data as possible.
- Allows correct use of ML algorithms — many need a full numeric matrix.
Reasons behind missing values
| Reason | Example |
|---|---|
| Technical issues | A sensor failed, data transmission dropped |
| Human errors | A field was skipped, wrong data entry |
| Privacy concerns | Respondents refuse to share income or health data |
| Data processing issues | Errors during merging, cleaning or transformation |
| Not applicable | SpouseName for an unmarried person |
Understanding the cause lets you judge whether the missingness is causing bias, and then choose the right method — imputation or removal.