Trimmed Mean Calculator
Calculate an average after removing a selected proportion of the lowest and highest observations. Compare the trimmed mean with the ordinary mean to see how extreme values affect your data.
Your trimmed mean
Enter a dataset and select a trim percentage, then select Calculate.
Show calculation
How to use this calculator
- Enter or paste your dataset, with observations separated as shown in the input instructions.
- Choose the percentage to remove from each end of the sorted values.
- Select Calculate to compare the trimmed mean with the ordinary mean and review the removed observations.
Trimmed mean formula
Sort n observations from smallest to largest. For a trim percentage p applied to each tail, remove k = floor(np/100) observations from the low end and k from the high end. The trimmed mean is the sum of the remaining n − 2k observations divided by n − 2k.
What the trim percentage means
A 10% trim removes up to 10% of the observations from each tail, not 10% total. Because an observation cannot be split, the count removed from each tail is rounded down. For small datasets, a chosen percentage may remove no values.
Worked example — outlier-resistant average
This example uses sample values, not your current inputs.
Consider 1, 2, 3, 4, 5, 6, 7, 8, 9, 100. The ordinary mean is 14.5. With a 10% trim from each tail, remove 1 and 100. The remaining values 2 through 9 have a trimmed mean of 5.5.
Frequently asked questions
What is a trimmed mean?
A trimmed mean is the arithmetic average after removing a specified number of the smallest and largest observations from a sorted dataset.
Does a 10% trimmed mean remove 10% in total?
No. In this calculator, 10% refers to each tail. For ten values, one low value and one high value are removed, so 20% of the observations are removed overall.
How many observations are removed?
The calculator multiplies the number of observations by the selected per-tail proportion and rounds the result down to a whole number for each tail.
What is the difference between trimming and winsorizing?
Trimming removes extreme observations before averaging. Winsorizing keeps the same number of observations but replaces extreme values with boundary values.
Is the trimmed mean always better than the ordinary mean?
No. Trimming reduces sensitivity to extreme values, but it also discards observations. The preferred summary depends on the data, study design and purpose.
Scientific foundations and academic references
The mathematical foundations, computational method and limitations of this tool.
Foundational research · 1977
Tukey, J. W. (1977). Exploratory Data Analysis. Addison-Wesley
John W. Tukey's Exploratory Data Analysis (1977) is a foundational work on resistant descriptive summaries and examining data beyond a single ordinary mean. It provides historical context for robust summaries, not a claim that this implementation reproduces a particular Tukey estimator.
Read source (opens in a new tab)Theoretical development · 1981
Huber, P. J. (1981). Robust Statistics. Wiley
Peter J. Huber's Robust Statistics (1981) formalized a broad mathematical framework for procedures that are less sensitive to contamination and outliers. A fixed-percentage trimmed mean is one simple resistant summary; it does not by itself perform robust inference.
Read source (opens in a new tab)Authoritative technical references
SciPy, scipy.stats.trim_mean
SciPy, scipy.stats.trim_mean
SciPy's trim_mean documentation specifies symmetric tail trimming, with the removed observation count rounded down. NIST's TRIMMED MEAN reference explains the motivation for discarding observations from the tails.
Read source (opens in a new tab)Computational method and verification
Numerical computation
The calculator sorts the dataset, removes floor(np/100) entries at each end and averages the remaining observations using IEEE 754 double-precision arithmetic. Sorting and trimming operate on unrounded values; the significant-digit control affects presentation only. This is a descriptive statistic, not an automatic outlier test or inferential confidence interval.
- 1, 2, 3, 4, 5, 6, 7, 8, 9, 100; 10% per tail
- One observation removed per tail; trimmed mean = 5.5
- 1, 2, 3, 100; 25% per tail
- Trimmed mean = 2.5
- 1, 2, 3, 100; 10% per tail
- No observations removed; mean = 26.5
Scope and limitations
This calculator sorts n observations and removes k = floor(np/100) observations from each tail, where p is the selected percent per tail. The mean is computed from the remaining n−2k observations. The rule matches SciPy trim_mean's floor convention. Ties are trimmed by sorted position, not by filtering out every copy of an equal boundary value. This is a descriptive summary, not an outlier hypothesis test.
Read full methodologyCite this tool
Use the following citation when referencing this calculator.
APA 7th edition
Statistics Calculator Lab. (2026). Trimmed Mean Calculator [Online calculator]. https://statscalculator.org/trimmed-mean-calculator/
Show BibTeX
@misc{statscalculator_trimmed_mean_2026,
author = {{Statistics Calculator Lab}},
title = {Trimmed Mean Calculator},
year = {2026},
howpublished = {Online calculator},
url = {https://statscalculator.org/trimmed-mean-calculator/}
}