SKEW in WPS Spreadsheet: Corrected Sample Skewness

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

A single numeric range may replace separate arguments. The original examples use nonconstant numeric samples containing at least three observations. They avoid argument-count limits and direct-versus-reference coercion differences.

For n observations with mean m and sample standard deviation s, the formula image gives n/((n−1)*(n−2))*sum(((x−m)/s)^3). Calculate s by dividing the squared-deviation sum by n−1 and then taking the square root. Substituting a population standard deviation changes the result.

Enter 1, 1, 4 in A2:A4 and use =SKEW(A2:A4) in C2.

1. Mean m=2; deviations are −1, −1, 2.
2. Squared deviations total 6; sample variance is 6/2=3, so s=sqrt(3).
3. Raw cubed deviations total −1−1+8=6.
4. Divide by s cubed and multiply by 3/(2*1).
5. The independent result is sqrt(3), approximately 1.73205080756888.

The relatively large upper observation makes the corrected sample statistic positive. The magnitude is dimensionless because deviations are divided by a standard deviation before cubing.

Adding a constant or multiplying by a positive scale leaves standardized shape unchanged. Multiplying by a negative scale reverses the sign, as the third row shows.

The symmetric three-point sample has zero skewness, but it is not evidence that the population is normal. Inspect the data distribution and sample size, and consider whether one extreme observation dominates the cubed-deviation sum. A statistic from three or four observations can be highly sensitive to a single value.

The shift and scale controls offer useful consistency checks when converting units. They do not justify deleting unusual observations. Any exclusion requires a substantive data-quality reason, not merely a desire to make skewness closer to zero.

The vendor’s nine values 110, 120, 124, 213, 214, 214, 128, 111, 124 produce the displayed 0.797646767, consistent with the corrected sample formula. The source also explicitly documents #DIV/0! for fewer than three points or zero sample standard deviation.

WPS distinguishes direct logical/numeric-text arguments from such values inside arrays or references. This guide does not generalize those coercion rules: all original observations are numeric cells, including a genuine zero in the final dataset. Error values and unconvertible text can cause errors according to the lesson, without a complete token table for every case.

See AVERAGE for the mean and DEVSQ for the squared-deviation component. Skewness is neither variance nor a test of normality by itself.

The official WPS lesson is dated 2019-06-12 and explicitly labels WPS Office for Windows Personal. Its numeric build is unspecified. This documents the function by that date; it does not establish first introduction, the last functioning build, other-platform parity or a retirement/deprecation history.

The complete lesson and its formula images and walkthrough were inspected on October 18, 2026. Original examples are documentation-derived independent calculations, not WPS or other spreadsheet-engine executions. Numeric reference values do not certify application precision or statistical model suitability.