Outlier sensitivity review
Compare statistics with flagged observations included and excluded, preserving every original row and exclusion decision.
Paste comma-separated CSV with a header and 2 to 2000 complete rows. Use dot decimals with at most six decimal places and values from -1000000000 to 1000000000. No missing values are silently dropped. Extra columns are used only to preserve the original input in reports. Calculations use the required columns only. Processing stays on your device. Results use floating-point arithmetic. Charts show the supplied sample, not proof about a population. Use a value column. Automatic flags use type-7 quartiles and 1.5 IQR fences. Optionally list original data-row numbers to exclude, separated by spaces. The comparison excludes automatic flags and your selected rows together, while retaining every original row in the reports. Empty or one-row retained sets have undefined sample variation. Flags do not prove errors; removing observations can bias conclusions. Search and sort change this comparison only. Copy result copies the summary. Download JSON or CSV for all observations, results, settings and method assumptions. Download SVG for the chart. One disposable worker per calculation, with a 10-second timeout. CSV supports up to 100 columns and 100000 cells through the shared parser. Paste comma-separated CSV with a header and 2 to 2000 complete rows. Use dot decimals with at most six decimal places and values from -1000000000 to 1000000000. No missing values are silently dropped. Extra columns are used only to preserve the original input in reports. Calculations use the required columns only. Processing stays on your device. Results use floating-point arithmetic. Charts show the supplied sample, not proof about a population. Use a value column. Automatic flags use type-7 quartiles and 1.5 IQR fences. Optionally list original data-row numbers to exclude, separated by spaces. The comparison excludes automatic flags and your selected rows together, while retaining every original row in the reports. Empty or one-row retained sets have undefined sample variation. Flags do not prove errors; removing observations can bias conclusions. Up to 1,000,000 characters. Review the settings before running.
Your result
How to use Outlier sensitivity review
- Paste complete numeric CSV rows with the required headers. Preserve the original order for lag diagnostics.
- Choose the displayed method settings and review their assumptions.
- Read the chart and table, then download all observations, results and assumptions as JSON or CSV.
Useful next steps
Outlier sensitivity review
Use a value column. Automatic flags use type-7 quartiles and 1.5 IQR fences. Optionally list original data-row numbers to exclude, separated by spaces. The comparison excludes automatic flags and your selected rows together, while retaining every original row in the reports. Empty or one-row retained sets have undefined sample variation. Flags do not prove errors; removing observations can bias conclusions.
Keep the complete data trail
Search and sort change this comparison only. Copy result copies the summary. Download JSON or CSV for all observations, results, settings and method assumptions. Download SVG for the chart. Group labels are preserved. Numeric CSV inputs use a decimal point; editable numeric settings also accept the selected language’s decimal notation.