CHISQ.TEST in Collabora Office: Compare Observed and Expected Counts
Use Collabora CHISQ.TEST on count arrays, reproduce the six-category p-value, audit chi-square contributions, and interpret the result without treating it as proof.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
Quick answer: For the vendor’s six observed counts 195,151,148,189,183,154 against expected counts 170 in each category, =CHISQ.TEST(A2:A7;B2:B7) returns a documented probability approximately 0.0209708029.
Syntax and the two arrays
=CHISQ.TEST(DataB; DataE)
DataB is the observed-count array and DataE is the corresponding expected-count range. Both are required. Keep the same category or table-cell ordering in the two arrays; swapping categories in only one changes the comparison.
The output is a probability associated with the chi-square comparison, not the chi-square statistic itself. In the independently checked source example it is the right-tail probability for the calculated statistic.
Use counts and a clearly specified expected model. An arbitrary list of percentages is not automatically an equivalent test, because sample size affects the statistic.
Example: reproduce the official six-category control
Enter A2:B7:
| Row | Observed | Expected |
|---|---|---|
| 2 | 195 | 170 |
| 3 | 151 | 170 |
| 4 | 148 | 170 |
| 5 | 189 | 170 |
| 6 | 183 | 170 |
| 7 | 154 | 170 |
The two totals are both 1020. In D2 enter =CHISQ.TEST(A2:A7;B2:B7). The documented result is approximately 0.0209708029, or 2.09708029%.
A small p-value is not the probability that the expected model is true or false. It summarizes how unusual the discrepancy is under the specified statistical comparison and its assumptions.
Independently reconstruct the statistic
In C2 enter =(A2-B2)^2/B2 and fill down through C7. The squared differences are 625,361,484,361,169,256, each divided by 170. Their sum is 2256/170, approximately 13.2705882353.
For this six-category fixed-expectation example, an independent chi-square survival calculation with 5 degrees of freedom gives approximately 0.0209708029, matching the vendor control. The statistic 13.2706 and probability 0.02097 are different outputs and should have different labels.
SUM can total the contributions. Do not use a left-tail cumulative probability as though it were the test’s right-tail result: its complement answers a different probability question.
Advanced example: construct expected counts for a 2×2 table
Use observed rows 30,20 and 10,40. Row totals are 50,50, column totals are 40,60, and the grand total is 100.
Under an independence model, each expected count is row total×column total/grand total. The expected rows are therefore 20,30 and 20,30. The independent statistic is approximately 16.666666667 and the chi-square right-tail probability with one degree of freedom is approximately 0.0000445571.
These are independently calculated mathematical expectations for the supplied 2×2 model, not a captured Collabora result. Put observed and expected tables in matching 2×2 ranges before passing them to CHISQ.TEST. Do not compare row totals alone when the question concerns association between the row and column classifications.
Advanced interpretation: sample size matters
If every count in the six-category example is doubled, each contribution (O−E)^2/E doubles, so the statistic becomes approximately 26.541176471. The category proportions are unchanged, but the evidence supplied by the larger count total is different.
This is why converting all counts to percentages and testing those percentages can give a misleading result. Preserve the sample counts and explain how expected counts were obtained.
Errors, assumptions, and compatibility
Expected counts in these calculations are positive; zero expected values make the displayed contribution formula divide by zero. Check matching dimensions, numeric counts, totals, and the validity of the expected model before interpreting the result. Small expected counts or dependent observations can undermine the approximation; the function’s successful calculation does not validate the study design.
The dedicated help does not comprehensively list native error codes, missing-data coercion, or every degrees-of-freedom adjustment for estimated model parameters. The controls here use fixed expectations or a transparent independence table.
The vendor identifies namespace COM.MICROSOFT.CHISQ.TEST, outside OpenDocument 1.3 OpenFormula. It does not establish full cross-product compatibility or identical behavior for every input.
Availability and verification
The exact function is documented in Collabora Office 26.04 desktop help, retrieved and reviewed on October 17, 2026. This establishes a documentation scope, not the first release that introduced the function or the last version in which it works; those versions and any retirement date remain unestablished. Collabora Online and mobile support are not inferred from desktop help, nor is behavior in another spreadsheet product.
The examples are documentation-based expectations with independent calculations. They have not been executed in Collabora Office. No application screenshot or native error/coercion certification is implied. Formulas use English function names and semicolon argument separators.
References
- Collabora Office 26.04: CHISQ.TEST (checked October 17, 2026).
- Editorial testing standards.