FTEST in WPS Spreadsheet: Compare Two Sample Variances

The official WPS lesson contains conflicting one-tail and two-tail wording. Its complete displayed example resolves the practical ambiguity: both samples have equal variance, and its result is 1, matching a two-tailed test rather than a one-tail probability of 0.5. The title and walkthrough also describe a two-tailed result.

The Open Platform FTest method independently describes two tails. It is JavaScript API documentation, not proof of identical availability across all worksheet clients. Together with the official worksheet demonstration, it is the primary support for this guide.

A separate WPS-hosted community/Q&A example supplies unequal-variance data and a numerical answer agreeing with the calculation below. The page explicitly says its user Q&A content has not been officially curated. Its Windows 10/WPS Office 2019 PC Personal label is a community-reported environment, not an official availability statement; the example is secondary numerical corroboration only.

Reviewed October 19, 2026. No exact introduction build, last-supported version, global-English, Mac, Linux, web or mobile availability was established. This guide uses source-demonstrated outputs and independent mathematical controls; no new native WPS execution or screenshot was performed.

=FTEST(array1,array2)

array1 and array2 are the two numeric sample arrays or ranges. In referenced arrays/ranges, WPS documents that text, logical values and blank cells are ignored; zero is included. Each sample needs at least two numeric observations and a nonzero variance. Otherwise the lesson documents #DIV/0!.

Keep labels outside the ranges. Two columns need not represent paired observations merely because they are next to each other. Decide whether the samples and the assumptions of a classical variance test make sense for the investigation before interpreting the formula.

Enter these two samples:

In D2 enter:

=FTEST(A2:A6,B2:B6)

The WPS-hosted community/Q&A example reports approximately 0.0039894255434. An independent numerical calculation gives 0.003989425543445509. These are consistent to the displayed precision.

For audit purposes, the ratio of sample variances is about 0.02670697909, with four degrees of freedom for each sample. The smaller one-tail probability is about 0.001994712772. Doubling that gives the reported result. This check uses the entire dataset, rather than selecting the source’s preferred wording.

Enter 1, 2, 3, 4, 5, 6 in A2:A7 and 7, 8, 9, 10, 11, 12 in B2:B7:

=FTEST(A2:A7,B2:B7)

Independent expected result: 1. Each sample variance is 3.5. The average levels differ, but shifting all observations by the same amount leaves their variance unchanged. This deliberately illustrates why FTEST is not a test of equal means.

The official lesson arranges those same values across two three-row, two-column blocks. It shows FTEST(B3:C5,B7:C9) and result 1. Flattening each block into one column preserves the numeric sample used by the independent calculation.

The classical F-test assumes independent random samples with independent observations from normal populations. It is sensitive to departures from normality. These are statistical assumptions, not conditions that the formula can establish for you; a mathematical demonstration does not verify the sampling assumptions. See the OpenStax explanation of the two-variance test.

Choose a significance threshold before looking at the result. A small p-value provides evidence against equal population variances under the test’s assumptions. It does not give the chance that the samples are identical, prove a cause, or quantify how important the difference is operationally.

Do not multiply the FTEST result by two again. It is already the two-tailed quantity in these verified WPS examples. Likewise, do not use the variance ratio itself as if it were the returned probability.

• #DIV/0!: count numeric cells in both ranges and check for a constant sample.
• Unexpected sample size: inspect numbers stored as text and blank cells; visible entries do not necessarily mean numeric observations.
• A result displayed as zero: show more decimal places or scientific notation before treating it as an exact zero.
• A result of 1 despite a large shift in values: compare spread separately from level, as in Example 2.
• An unexpected result after copying: inspect relative references and ensure each formula still points to the intended samples.

These examples cover ordinary numeric ranges. They do not establish every scalar-coercion, array-error or cross-version compatibility behavior.

Use VAR or VAR.S to inspect the sample variances separately. FDIST calculates a right-tail probability from an F statistic and degrees of freedom; it does not take the two raw sample ranges used by FTEST.