FISHER in Collabora Office: Transform a Correlation-Scale Value
Understand Collabora FISHER, verify its logarithmic transform and inverse, and avoid treating transformed values as probabilities.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
Quick answer: =FISHER(0.5) has independently expected result approximately 0.549306144. It transforms a value strictly between −1 and 1 onto an unbounded real-number scale.
Syntax and mathematical meaning
=FISHER(Number)
Number is the value to transform; it is the function’s only required argument. The dedicated Collabora help calls this the Fisher transformation and gives a rounded result of 0.55 for input 0.5.
The independent mathematical definition is 0.5*LN((1+x)/(1-x)), also called the inverse hyperbolic tangent. The transform is increasing and odd: changing the sign of x changes the sign of the result. It does not turn an arbitrary dataset into normal data merely because the help refers to a distribution close to normal.
For ordinary finite real-valued use, require −1<x<1. This domain is a mathematical restriction of the logarithmic expression. The guide does not claim a particular native error display at either endpoint.
Example 1: create a transparent transform table
Put −0.8, −0.5, 0, 0.5, 0.8 in A2:A6. In B2 enter =FISHER(A2) and fill down. Independent expected results are approximately −1.098612289, −0.549306144, 0, 0.549306144, 1.098612289.
In C2 enter =0.5*LN((1+A2)/(1-A2)) and fill down. This is a formula-level mathematical cross-check, not independent application execution. In D2 use =B2-C2; the target is zero apart from numerical rounding.
The transformed value for 0.8 exceeds one. That is legitimate because a Fisher-transformed value is not a correlation or a probability. Do not apply a percentage format and interpret 1.0986 as a probability above 100%.
Example 2: reverse the transformation
In E2 enter =FISHERINV(B2) and fill down. The independent round-trip targets are the original A2:A6 values. Compare the stored values rather than manually retyping a three-decimal display.
For example, reversing the full transform of 0.8 targets 0.8, whereas reversing a deliberately rounded 1.099 targets a slightly different value. The lost precision is a consequence of rounding the intermediate number, not a failure of the inverse relationship.
Advanced application: show how the scale stretches
The change from x=0 to x=0.5 becomes approximately 0.549306144 on the transformed scale. The change from x=0.5 to x=0.8 becomes approximately 0.549306144 again, even though its original-scale width is only 0.3.
Near either endpoint, the transformation stretches small changes increasingly strongly. This matters when displaying transformed values beside original correlation-scale values: equal distances in one column do not imply equal distances in the other.
Use separate column labels such as “Original value” and “Fisher-transformed value.” Copying the same axis label onto both columns obscures their different units and domains.
Advanced application: averaging is scale-dependent
The arithmetic average of 0.2 and 0.8 is 0.5. Transforming both, averaging the transforms and reversing gives approximately 0.572122462.
An exact worksheet layout is A10=0.2 and A11=0.8; B10==FISHER(A10) and B11==FISHER(A11); B13==FISHERINV(AVERAGE(B10:B11)). The different result demonstrates nonlinearity. It is not a recommendation for pooling correlations: a valid statistical combination also depends on study structure, sample sizes and assumptions.
Errors and troubleshooting
Validate numeric inputs strictly inside the open interval. Values exactly ±1 make the transform diverge mathematically; values outside the interval are outside this real-valued definition. The documented example does not establish native error codes or all extreme-value behavior.
If a round trip is inaccurate, inspect early rounding and proximity to an endpoint. If results appear as percentages, change the display to an ordinary number and revisit the column label. No native Collabora execution or statistical-normality validation was performed.
Availability and verification
The exact function is documented in Collabora Office 26.04 desktop help. Its source snapshot was retrieved on 2026-10-13 and hash-verified for this guide on October 15, 2026. Desktop documentation is not proof of Online or mobile support.
First introduced release and last functioning version are unknown. No native Collabora or WPS run occurred. The numerical values are independent mathematical expectations; they do not certify application-specific numerical tolerances, error codes or every edge case. Formulas use English names and semicolon separators.