Statistical analysis service
Data Cleaning and Statistical Diagnostics
Prepare an auditable analytical dataset and review missingness, coding, outliers and model-specific assumptions.
Request statistical guidanceReliable analysis starts before modelling
Data cleaning should be documented rather than performed as an invisible preliminary step. Coding errors, impossible values, duplicate records and inconsistent missing-value conventions can materially change results.
Checks can include
- variable types and labels
- range and logic checks
- duplicate observations
- missingness patterns
- outliers and influential cases
- distributional and model-specific assumptions
Keep a reproducible record
Where possible, cleaning decisions should be scripted or logged so that the analytical dataset can be rebuilt from the original source without relying on memory.
Request guidance for this analysis
Describe the study context so the method, assumptions and scope can be assessed before any commitment is made.
