About the MAD Calculator
The Median Absolute Deviation calculator measures how spread out a dataset is using medians instead of means, which keeps a handful of extreme values from dominating the result. It suits datasets like household incomes, response times, or sensor readings, where one or two extreme entries would otherwise inflate a standard deviation calculation.
How It Works
Enter a list of numbers, and the calculator first finds the median of the raw data. It then computes the absolute difference between every value and that median, and takes the median of those differences; that final number is the MAD. Both median steps use the same rule: sort the values and take the middle one, or average the two middle values when the count is even.
Formula & Methodology
To compute it by hand: sort the data and find the median (the middle value, or the average of the two middle values for an even count). Subtract that median from every data point and take the absolute value of each result. Sort those absolute deviations and find their median using the same middle-value rule; that is the MAD.
Examples
Skewed dataset with an outlier
For 4, 6, 7, 7, 8, 9, 10, 45, the median is 7.5. The absolute deviations are 3.5, 1.5, 0.5, 0.5, 0.5, 1.5, 2.5, and 37.5, and the median of those is 1.5, so MAD = 1.5, barely nudged by the 45 that would have wrecked a standard deviation calculation.
Small symmetric-ish dataset
For 2, 3, 3, 4, 5, 7, 10, the median is 4. The absolute deviations sorted are 0, 1, 1, 1, 2, 3, 6, and their median is 1, giving MAD = 1.
Advantages
- Resistant to the influence of a small number of extreme values, unlike standard deviation, which squares every deviation and lets outliers dominate.
- Works directly from raw values with no distributional assumptions, so it applies equally well to skewed or heavy-tailed data.
- Doubles as the building block for outlier-flagging methods, such as the modified z-score, that need a spread measure not corrupted by the outliers themselves.
Common Mistakes
- Comparing a raw MAD value directly against a standard deviation figure without rescaling, since the two are not on the same numeric scale by default.
- Assuming MAD behaves like a percentage or bounded score; it is expressed in the same units as the original data, not a normalized measure.
- Forgetting that outliers still show up in the sorted absolute-deviation list; MAD reduces their influence, it doesn't discard them.
Edge Cases to Watch For
- At least one number must be entered; an empty input returns an error.
- Even-length lists average the two middle values twice, once for the data median and once again for the MAD, since both steps use the identical median function.
- A MAD of exactly 0 is a valid output whenever more than half the values are identical, and it signals that outlier-detection formulas built on top of MAD, which divide by it, will break down.
- MAD sits on a different scale than standard deviation by default; multiplying it by roughly 1.4826 approximates the standard deviation for normally distributed data.
Common Use Cases
- Analysts summarizing spread in datasets known to contain extreme values, such as transaction amounts or response times.
- Quality control processes that need a spread estimate unaffected by a handful of defective or extreme measurements.
- Anyone preparing inputs for a robust outlier-detection method that requires MAD as a preliminary step.