Outlier Detector
Paste your data to flag unusual values. IQR is the most robust choice, Z-score suits roughly normal data, and MAD resists distortion by the outliers themselves.
Method:
Threshold:
Output:
Separator:
Statistics Result
Download CSV
| No. | Original | Outlier? | Deviation Score |
|---|
Introduction to the tool and how to use it
An outlier is a value that clearly stands apart from the rest of the data — it may be a typo, or it may be the most interesting number of the lot. This tool pulls them out with three industry-standard methods.
The three methods:
· IQR — the most robust and most common. Work out the 25th and 75th percentiles; their difference is the IQR. Anything above Q3 + threshold×IQR or below Q1 − threshold×IQR is flagged (threshold defaults to 1.5; use 3 for extreme values only);
· Z-score — how many standard deviations a value sits from the mean; good for roughly normal data (thresholds of 2 or 3 are typical);
· MAD — measures spread using the median of the absolute deviations from the median, so the outliers themselves distort it far less than they do the standard deviation (thresholds of 2.5 or 3.5 are typical).
Steps:
1. Paste the data; commas, spaces, newlines or any separator work;
2. Pick the Method;
3. Set the Threshold — smaller is more sensitive and flags more values;
4. Choose what to output: list everything (each row marked as outlier or not) / only outliers;
5. Set the separator to control the text box layout;
6. Click Run. Results appear in the text box and in the table below (index / value / outlier? / deviation score).
Note: the Deviation Score means different things per method — for IQR it is how many IQRs beyond the fence the value sits (1.5 means it is exactly on the threshold), for Z-score it is |z|, and for MAD it is the distance in units of 1.4826×MAD. Bigger means more extreme, so you can sort by this column for a second pass. Values are always listed in input order so you can line them up with the original data; when the spread is 0 (all values identical) nothing is flagged as an outlier.
The three methods:
· IQR — the most robust and most common. Work out the 25th and 75th percentiles; their difference is the IQR. Anything above Q3 + threshold×IQR or below Q1 − threshold×IQR is flagged (threshold defaults to 1.5; use 3 for extreme values only);
· Z-score — how many standard deviations a value sits from the mean; good for roughly normal data (thresholds of 2 or 3 are typical);
· MAD — measures spread using the median of the absolute deviations from the median, so the outliers themselves distort it far less than they do the standard deviation (thresholds of 2.5 or 3.5 are typical).
Steps:
1. Paste the data; commas, spaces, newlines or any separator work;
2. Pick the Method;
3. Set the Threshold — smaller is more sensitive and flags more values;
4. Choose what to output: list everything (each row marked as outlier or not) / only outliers;
5. Set the separator to control the text box layout;
6. Click Run. Results appear in the text box and in the table below (index / value / outlier? / deviation score).
Note: the Deviation Score means different things per method — for IQR it is how many IQRs beyond the fence the value sits (1.5 means it is exactly on the threshold), for Z-score it is |z|, and for MAD it is the distance in units of 1.4826×MAD. Bigger means more extreme, so you can sort by this column for a second pass. Values are always listed in input order so you can line them up with the original data; when the spread is 0 (all values identical) nothing is flagged as an outlier.
Message board
All messages →-
No one has spoken up yet — want to go first?