AVEDEV in Excel: Average Absolute Deviations From the Mean

• Introduced: first introduction version is not established by the official history reviewed; the current product list is not a launch record.
• Availability: official help lists Excel for Microsoft 365, Excel for Microsoft 365 for Mac, Excel 2024, Excel 2024 for Mac, Excel 2021, Excel 2021 for Mac, Excel 2019, Excel 2016.
• Checked: October 14, 2026; official AVEDEV reference. Documentation and independently calculated examples; no native Excel execution or broader platform certification.

=AVEDEV(number1, [number2], ...)

The first argument is required, with up to 255 arguments. An argument can be a numeric value or a range. Microsoft distinguishes direct logical/numeric-text arguments, which count, from text, logical values and empty cells inside referenced arrays, which are ignored. Numeric zero participates.

Use English function names and comma separators for these controls. The measure uses absolute deviations around the mean, not around the median, and its unit matches the input measurement unit.

Enter A2:A5 as 2, 4, 6, 8. In B2 use and fill down:

=ABS(A2-AVERAGE($A$2:$A$5))

Expected B2:B5 values are 3, 1, 1, 3. The total absolute deviation is eight across four observations:

=AVEDEV(A2:A5)
=AVERAGE(B2:B5)

Both expected results are 2. The fixed mean range stays A2:A5 when the helper formula is filled. Letting that range drift would compare each row with a different population.

With numeric 2, 4, 6 in D2:D4, =AVEDEV(D2:D4) has expected result 1.333333333. Their mean is four and the distances are two, zero and two.

Add numeric 0 in D5 and extend the range. The mean becomes three and distances become one, one, three and three, giving AVEDEV 2. A zero changes both the mean and the set of distances; it is not discarded as a missing reading.

Do not substitute a visually blank or formula-empty value for the explicit zero and assume the same participant count. The documented referenced-empty rule and every imported-type edge are separate questions.

Create F2:F4 with numeric 2, text 4 entered with a leading apostrophe, and formula =TRUE():

=AVEDEV(F2:F4)
=AVEDEV(2,"4",TRUE)

Expected results under Microsoft’s stated participation rules are 0 and approximately 1.111111111. The referenced control has one numeric participant, two, whose distance from its own mean is zero. The direct control counts 2, 4 and 1, with mean 7/3 and total absolute deviation 10/3 across three values.

This is a narrowly documented type contrast. It does not certify every nonnumeric direct string or every error-containing range.

For the original four values, adding ten to every observation produces 12, 14, 16, 18 and leaves AVEDEV at 2. Each value and the mean move together, preserving the distances.

Multiplying every original value by negative three produces -6, -12, -18, -24 and gives AVEDEV 6. Reversing the sign does not make distances negative; their magnitude scales by three.

These checks help detect an unintended change of units. An average absolute deviation of two meters and two centimeters represents different physical variability even though the displayed number matches.

For 2, 4, 6, 20, the arithmetic mean is eight. The absolute distances are 6, 4, 2, 12, so AVEDEV is 6.

The median of the same observations is five. Their distances from that median are 3, 1, 1, 15; the median of those distances is 2. That latter statistic is a median absolute deviation, not what AVEDEV calculates.

Both measures can be meaningful, but the choice must match the question. See AVERAGE, MEDIAN and ABS for the component operations.

Keep the source population fixed, inspect actual zeros and stored types, and state the measurement unit. Do not confuse AVEDEV with the square-root-based standard deviation or with DEVSQ, which sums squared deviations.

The selected reference does not supply a complete exact-error table. No native last-digit accuracy, extreme-input behavior or automatic cleaning of error cells is claimed by these finite controls.

• Microsoft AVEDEV reference.
• Editorial testing standards.

Inputs are fictional. Decimal answers are independent mathematical expectations rounded to nine places, not captured native results. Extreme magnitudes, complete coercion matrices and cross-version numerical accuracy remain untested.