CONFIDENCE Function (OpenOffice Calc)

• Introduced: First release not established. The OpenOffice.org 3.3 appendix documents the name; that is evidence of presence, not its first introduction.
• Supported in: Apache OpenOffice 4.1.16; no final supported release established.
• Scope: Desktop Calc. Argument contracts and boundaries below refer to the cited release.
• Checked: October 14, 2026. Documentation/source and independent mathematics checked; Apache OpenOffice Calc was not run for this guide.

CONFIDENCE uses a two-sided normal critical value. For significance alpha, known population standard deviation sigma and sample size n, its margin is z(1−alpha/2)×sigma/SQRT(n). Alpha=0.05 corresponds to a 95% confidence procedure. Function reference.

The standard deviation describes the population measurements, while the divisor uses the number of sampled observations. The OpenOffice reference correctly describes Size as the sample size.

=CONFIDENCE(Alpha; StDev; Size)

All three arguments are required. The release evaluator requires 0<Alpha<1, StDev>0 and an integer-truncated Size of at least 1. Fractional sample counts are floored, but a real sample count should already be an integer.

Enter 5% or 0.05 for a 95% confidence level, not 95%. Alpha is the omitted tail probability across both tails, not the retained confidence percentage. Keep StDev in the same units as the sample mean so the resulting margin has those units too.

Expected values below are independently calculated and rounded to twelve decimal places.

Suppose the sample mean is 50, the known population SD is 12, and 100 independent observations were sampled. Put 0.05, 12, 100 and 50 in A2:D2. In E2 enter =CONFIDENCE(A2;B2;C2); then F2==D2-E2 and G2==D2+E2.

The margin is 2.351956781448, and the endpoints are approximately 47.648043218552 and 52.351956781448. Report the assumptions with the result. A narrow interval cannot correct a biased sampling method or dependence that makes the effective sample size smaller.

With the same known SD, increasing n from 100 to 400 halves the margin to 1.175978390724. Precision improves with the square root of sample size, so doubling the sample does not halve the margin.

For a target margin no greater than 2 at 95% confidence, the planning equation is n≥(z(0.975)×12/2)². Its value is approximately 138.292517544989, so round up to 139 observations. Using 138 would slightly exceed the target margin. This calculation assumes the same model and known standard deviation.

A 95% confidence procedure is about the long-run coverage of intervals from repeated samples. It does not mean that 95% of individual measurements are inside the reported interval. A prediction interval answers that different question.

If sigma is unknown and estimated from the sample, blindly substituting a sample SD into the known-sigma calculation is not an exact normal-population confidence procedure. A Student t method is commonly appropriate under the corresponding independent normal-sample assumptions. Choose the statistical method before choosing the formula.

This guide checks ordinary numerical examples and source-level input guards. It does not validate a particular dataset, sampling design or extremely small alpha on every installed release.

• CONFIDENCE documentation
• Apache OpenOffice 4.1.16 evaluator and argument contract