ONLYOFFICE FISHERINV: Inverse Transform and Precision Examples
Reverse the Fisher transformation in ONLYOFFICE, compare exponential calculations and preserve precision through reproducible round-trip examples.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
FISHERINV converts a Fisher-scale value back to the original −1…1 scale. It is an inverse transformation, not a reciprocal, and it does not require the input itself to lie between −1 and 1.
Availability and version history
- Introduced: The first introduction release and date are not established by the checked sources.
- Availability: Currently documented in ONLYOFFICE’s rolling function help; no final supported-version endpoint is established.
- Scope: Shared Spreadsheet Editor documentation, with formulas corroborated by a pinned official source-code revision. Individual client/build parity is unverified.
- Checked: October 14, 2026, against the official FISHERINV reference, source and vendor screenshot. Example outputs were independently calculated; these cases were not tested in the native application.
Syntax and mathematical meaning
=FISHERINV(y)
The official reference requires one numeric value. It does not document optional arguments or a sample-size parameter.
For ordinary finite mathematical inputs, the inverse is:
=(EXP(2*y)-1)/(EXP(2*y)+1)
The official source at a pinned revision uses that expression. Its mathematical output lies strictly between −1 and 1, although finite-precision display and extreme-input handling need separate consideration. This source snapshot is not a mapping to every installed Docs or Desktop build.
Example: positive and negative transformed inputs
Create InverseScale with Fisher value, Inverse result and Exponential check in A1:C1. Enter the inputs shown below in A2:A8.
| Row | A: y | Expected B and C, approximately |
|---|---|---|
| 2 | -2 | -0.9640275801 |
| 3 | -1 | -0.7615941560 |
| 4 | -0.5 | -0.4621171573 |
| 5 | 0 | 0 |
| 6 | 0.5 | 0.4621171573 |
| 7 | 1 | 0.7615941560 |
| 8 | 2 | 0.9640275801 |
Enter =FISHERINV(A2) in B2 and =(EXP(2*A2)-1)/(EXP(2*A2)+1) in C2. Fill B2:C2 through row 8.
Unlike FISHER’s input, the transformed values −2 and 2 are not outside the documented inverse function’s numeric-input description. As the magnitude grows, the inverse approaches an endpoint rather than growing without limit. Negative outputs also show why this is not a probability calculation.
Put Forward check in D1, enter =FISHER(B2) in D2 and fill through D8. For these moderate inputs, the mathematical targets are −2,−1,−0.5,0,0.5,1,2. Use full stored B values for this check rather than retyping rounded display text.
Inverse does not mean reciprocal
On InverseScale, put =1/A7 in F2 and =FISHERINV(A7) in F3. Since A7 is 1, the first result is 1, while the inverse Fisher result is approximately 0.7615941560.
The inverse reverses FISHER’s logarithmic transformation; it does not calculate one divided by the transformed value. At y = 0, FISHERINV gives 0 mathematically, while a reciprocal would be undefined.
Preserve precision in a round trip
Create RoundTrip. Enter 0.2, 0.6 and 0.8 in A2:A4, with Original, Fisher value and Recovered value in A1:C1. Enter =FISHER(A2) in B2 and =FISHERINV(B2) in C2; fill through row 4. The mathematical C targets are 0.2,0.6,0.8.
The transformed value for 0.8 is approximately 1.0986122887. To examine a deliberately rounded intermediate value without overwriting it, enter the number 1.1 in E2. In F2 enter =FISHERINV(E2); expect approximately 0.8004990218. In F3 enter =F2-A4; expect approximately 0.0004990218.
Formatting B4 to show two decimals does not ordinarily replace its stored value with 1.1. Manually retyping that displayed number does. Keep the calculation chain linked when accuracy matters.
Vendor examples
Create VendorInverse with 0.56, −0.89 and 0.12 in A1:A3. Enter =FISHERINV(A1) in C1 and fill to C3.
| Input | Independent value, approximately | Vendor display |
|---|---|---|
| 0.56 | 0.5079774329 | 0.5079774 |
| -0.89 | -0.7113937318 | -0.7113937 |
| 0.12 | 0.1194272985 | 0.1194273 |
The values agree with the official illustration at its displayed precision. The same three numbers produce different results when supplied to FISHER; the forward and inverse functions are not interchangeable.
Numerical and statistical limits
The exponential expression can become numerically difficult for very large positive inputs, and values close to ±1 can appear rounded to an endpoint. The short help does not publish a safe magnitude threshold or an exact error result for every extreme. No such threshold is inferred from the source snapshot.
These examples do not establish native text conversion, imported-array behavior or cross-client precision. More importantly, the inverse transformation does not by itself create a valid confidence interval or combined correlation estimate. Those uses need a separately justified statistical model.