What should agree in an fsQCA comparison?
Fuzzy-set qualitative comparative analysis, or fsQCA, looks for combinations of conditions associated with an outcome. A membership value describes how strongly a case belongs to a set. A truth table groups cases by their combinations of conditions, and the analysis simplifies those combinations into solution terms.
A useful comparison must follow the same cases and memberships through those steps. It should check which truth-table rows are included, which alternative solutions are found, and how well those solutions fit the data. Matching the printed formula alone is not enough: equivalent logical expressions can be written differently.
Consistency measures how closely the memberships follow the set relationship being tested. Coverage describes how much of the relevant set is covered by that relationship. Necessity and sufficiency ask different questions, so their consistency and coverage values must be compared under the same definitions.
Which software produced these results?
The rows below identify the QuickPLS result and reference used in each test. The independent enumerator works through possible logical combinations without using the R QCA package. The other rows name QCA where that package supplied a reference result. These are the versions used for the saved comparisons, not a list of the current product releases.
On a small screen, swipe the table sideways to see every column.
| Record | QuickPLS build or result checked | Reference | Comparison outcome |
|---|---|---|---|
| Original fsQCA independent mathematics | Saved fsQCA calculation test case; reviewed 28 October 2026 | Independent ternary-cube enumerator | Passed the comparisons documented for this record |
| QCA metric signal | Saved fsQCA calculation test case; reviewed 28 October 2026 | QCA 3.25 | Passed the comparisons documented for this record |
| QCA truth table/minimization | Saved fsQCA calculation test case; reviewed 28 October 2026 | QCA 3.25 | Passed the comparisons documented for this record |
| Installed fsQCA NF | 1.01.19 | QCA 3.25, independent unique-coverage math | Passed the comparisons documented for this record |
An independent calculation of the saved fsQCA result
The reference checked one saved QuickPLS output containing 410 decimal values, integer values and undefined results. It separately worked through the possible combinations of conditions to check the logic, rather than reusing the QuickPLS calculation. The numerical results in this example had no reported difference.
- This is an independently calculated reference. It does not run the R QCA package.
- The exhaustive reference works through models with at most six conditions. Its ability to check every combination within that limit does not establish the same coverage for larger models.
| What was checked | Value / scope |
|---|---|
| Compared numbers and missing-value statuses | 410 |
| Outside the stated tolerance | 0 |
| Different scenario names in the comparison record | 1 |
| Largest absolute difference | 0 |
Selected values from the independent calculation
For example, the first row reports consistency of 0.8 in both calculations. The rows also include a chosen consistency threshold and necessity measures for individual truth-table rows. A threshold is an input to the analysis, whereas a calculated consistency value is a result; both can be checked, but they answer different questions.
The table displays values to 15 significant digits. Differences were calculated before this display rounding. Use the download for the original recorded precision and any adjustments made to compare the same definition of a result. Expand Result identifier to see the exact field in the saved output.
On a small screen, swipe the table sideways to see every column.
| Quantity compared | QuickPLS | Reference | Absolute difference | Comparison result |
|---|---|---|---|---|
Sufficiency consistency (set-fit example)Result identifier/metric_signal/consistency/value | 0.8 | 0.8 | 0 | Pass |
Necessity consistency (set-fit example)Result identifier/metric_signal/necessity_consistency/value | 0.8 | 0.8 | 0 | Pass |
Truth-table consistency cutoffResult identifier/nf/thresholds/consistency | 0.8 | 0.8 | 0 | Pass |
Sufficiency consistency, truth-table result row 1Result identifier/nf/rows/0/fit/consistency/value | 1 | 1 | 0 | Pass |
Necessity consistency, truth-table result row 1Result identifier/nf/rows/0/fit/necessity_consistency/value | 0.176470588235294 | 0.176470588235294 | 0 | Pass |
Sufficiency consistency, truth-table result row 2Result identifier/nf/rows/1/fit/consistency/value | 0.736842105263158 | 0.736842105263158 | 0 | Pass |
Selected set-fit measures compared with QCA
This smaller test compares selected consistency and coverage measures with QCA. It asks whether the two calculations give the same set-fit measures for this example; it does not test the complete route from raw data to a final solution. The table reports the number of comparisons and the largest numerical difference.
- The result supports agreement for the listed measures in this example. It does not establish agreement for calibration, every truth-table decision or every type of solution.
| What was checked | Value / scope |
|---|---|
| Compared numbers and missing-value statuses | 6 |
| Outside the stated tolerance | 0 |
| Different scenario names in the comparison record | 1 |
| Largest absolute difference | 0 |
Consistency values from the set-fit example
The displayed consistency and necessity-consistency values are both 0.8 in QuickPLS and the reference. They happen to be equal in this example, but the two labels describe different set relationships and should not be treated as interchangeable measures.
Values are displayed to 15 significant digits. The difference column uses the values before that display rounding; the download preserves their recorded precision and identifies any adjustment needed to compare the same definition.
On a small screen, swipe the table sideways to see every column.
| Quantity compared | QuickPLS | Reference | Absolute difference | Comparison result |
|---|---|---|---|---|
Sufficiency consistencyResult identifierconsistency | 0.8 | 0.8 | 0 | Pass |
Necessity consistencyResult identifiernecessity_consistency | 0.8 | 0.8 | 0 | Pass |
Truth-table rows and simplified solutions compared with QCA
This comparison covers the NF truth table and minimum conservative and parsimonious solutions for the LF/LF-multiple examples. It also includes selected intermediate solutions. These names identify the saved examples, so use them when finding the corresponding records in the download.
Minimisation means finding simpler combinations of conditions that cover the selected configurations. Conservative solutions avoid using unobserved configurations to simplify the result. Parsimonious solutions allow those simplifying assumptions more broadly, while intermediate solutions use selected assumptions. Comparing these solution types therefore requires matching their assumptions as well as their final expressions.
- Permission to redistribute the published datasets has not been established, so this download does not supply those datasets.
- The comparison has no corresponding QuickPLS calibration output. It does not show that the raw data were converted to memberships identically.
- For LF-multiple, separate mathematical calculations check the intermediate-solution steps. Do not read those steps as an additional direct QCA package comparison.
| What was checked | Value / scope |
|---|---|
| Compared numbers and missing-value statuses | 142 |
| Outside the stated tolerance | 0 |
| Different scenario names in the comparison record | 6 |
| Largest absolute difference | 2.22044604925031e-15 |
Selected truth-table values
The row labels identify the configuration and the measure being compared. For example, Sufficiency consistency, configuration 0000 agrees in both calculations. Another row has a very small reported difference even though the displayed values look equal. Expand Result identifier to match a row to its original field in the download.
The table uses 15 significant digits. The difference was calculated before that display rounding. The download retains the original precision and explains any adjustments needed to put the two outputs on the same definition.
On a small screen, swipe the table sideways to see every column.
| Quantity compared | QuickPLS | Reference | Absolute difference | Comparison result |
|---|---|---|---|---|
Sufficiency consistency, configuration 0000Result identifier0000/consistency | 1 | 1 | 0 | Pass |
Necessity consistency, configuration 0000Result identifier0000/necessity_consistency | 0.176470588235294 | 0.176470588235294 | 0 | Pass |
Sufficiency consistency, configuration 0001Result identifier0001/consistency | 0.736842105263158 | 0.736842105263158 | 1.11022302462516e-16 | Pass |
Necessity consistency, configuration 0001Result identifier0001/necessity_consistency | 0.205882352941176 | 0.205882352941176 | 0 | Pass |
Sufficiency consistency, configuration 0010Result identifier0010/consistency | 1 | 1 | 0 | Pass |
Necessity consistency, configuration 0010Result identifier0010/necessity_consistency | 0.205882352941176 | 0.205882352941176 | 0 | Pass |
The NF example calculated in the installed app
This test used existing memberships for 12 NF cases. It checked the installed app output against QCA and an independent calculation of unique coverage, the contribution of a solution term that is not covered by the other terms. There were 539 checks of numerical or missing results and 365 checks of output structure, such as whether the expected parts of the result were present.
- The count includes 128 comparisons reused from earlier checks. They are part of the reported total, not additional independent runs.
- Three resource counters, which describe computation rather than research findings, have no matching reference.
- This example excludes calibration, intermediate solutions, missing values, crossover cases and models with many conditions. A crossover case has membership exactly at the point between being more in and more out of a fuzzy set.
- Permission to redistribute the NF data has not been established.
| What was checked | Value / scope |
|---|---|
| Compared numbers and missing-value statuses | 539 |
| Outside the stated tolerance | 0 |
| Different scenario names in the comparison record | 1 |
| Largest absolute difference | 2.22044604925031e-15 |
Selected condition-fit values from the installed app
Each pair of rows compares sufficiency consistency with necessity consistency for the same set. The displayed labels identify A present, A absent and I present. Expand Result identifier to find the corresponding field in the download; keep the presence or absence of the condition with each measure.
The displayed values use 15 significant digits. The difference column was calculated before display rounding, so a small nonzero difference may remain when the two values look the same. The download preserves the recorded precision and any adjustments used to match the definitions.
On a small screen, swipe the table sideways to see every column.
| Quantity compared | QuickPLS | Reference | Absolute difference | Comparison result |
|---|---|---|---|---|
Sufficiency consistency: A presentResult identifierfsqca_condition_fit/0/consistency | 0.654320987654321 | 0.654320987654321 | 0 | Pass |
Necessity consistency: A presentResult identifierfsqca_condition_fit/0/necessity_consistency | 0.779411764705882 | 0.779411764705882 | 3.33066907387547e-16 | Pass |
Sufficiency consistency: A absentResult identifierfsqca_condition_fit/1/consistency | 0.846153846153846 | 0.846153846153846 | 1.11022302462516e-16 | Pass |
Necessity consistency: A absentResult identifierfsqca_condition_fit/1/necessity_consistency | 0.485294117647059 | 0.485294117647059 | 1.11022302462516e-16 | Pass |
Sufficiency consistency: I presentResult identifierfsqca_condition_fit/2/consistency | 0.757142857142857 | 0.757142857142857 | 1.11022302462516e-16 | Pass |
Necessity consistency: I presentResult identifierfsqca_condition_fit/2/necessity_consistency | 0.779411764705882 | 0.779411764705882 | 3.33066907387547e-16 | Pass |
The error limits for the independent calculation
For decimal values, the allowed absolute difference is <= 1e-12. Integer values and missing-result positions must match exactly. Logic and result-structure checks are reported separately.
The table distinguishes decimal calculations, integer counts and results that are undefined. A missing result must occur in the same position in both outputs; it cannot be replaced with a numerical answer. These counts describe individual values or missing-result positions within the saved example. Logical and structural checks are counted separately.
On a small screen, swipe the table sideways to see every column.
| Absolute tolerance | Relative tolerance | Numeric or missing-value checks | Where the comparison rule was specified |
|---|---|---|---|
| 1e-12 | 0 | 104 | Comparison script |
| 0 | 0 | 303 | Comparison script |
| Unavailable | Unavailable | 3 | Missing-value status only |
The error limits for the selected set-fit measures
abs(native - aligned_reference) <= absolute_tolerance + relative_tolerance * abs(aligned_reference)
Here native means the QuickPLS value, and aligned_reference means the reference expressed under the same definition. The allowed difference combines a fixed allowance with an allowance proportional to the reference value. The table records both. A missing result must also agree with the reference; the comparison count is a count of values, not of separate studies.
On a small screen, swipe the table sideways to see every column.
| Absolute tolerance | Relative tolerance | Numeric or missing-value checks | Where the comparison rule was specified |
|---|---|---|---|
| 1e-08 | 1e-08 | 6 | Comparison script |
The error limits for the truth-table comparison
abs(native - aligned_reference) <= absolute_tolerance + relative_tolerance * abs(aligned_reference)
The formula compares each QuickPLS value with the corresponding reference after their definitions have been matched. It allows the fixed difference and proportional difference shown in the table. A result expected to be missing must be missing in both outputs. The reported checks come from the saved examples, rather than from that many separate studies.
On a small screen, swipe the table sideways to see every column.
| Absolute tolerance | Relative tolerance | Numeric or missing-value checks | Where the comparison rule was specified |
|---|---|---|---|
| 1e-08 | 1e-08 | 142 | Comparison script |
The error limits for the installed app example
abs(native - aligned_reference) <= absolute_tolerance + relative_tolerance * abs(aligned_reference)
The table separates results that must agree exactly from decimal results that allow a small difference. In the formula, native is the QuickPLS value and aligned_reference is the reference under the same definition. The allowance has a fixed part and a part proportional to the reference value. Missing-result positions must agree as well. Each comparison concerns an output value or missing result, not a separate study.
On a small screen, swipe the table sideways to see every column.
| Absolute tolerance | Relative tolerance | Numeric or missing-value checks | Where the comparison rule was specified |
|---|---|---|---|
| 0 | 0 | 176 | Specified separately for each compared value |
| 1e-08 | 1e-08 | 363 | Specified separately for each compared value |
What these comparisons let you conclude
The results support the specific calculations listed above: independent logical enumeration, selected QCA measures and solutions, and the installed app example. A check is one value or output property; many checks can come from the same dataset and analysis.
Use the result table and its error limits together. A zero difference in the first example is not a claim that every comparison on this page has a zero difference. The downloads show the full recorded precision behind the rounded display.
- Independent enumeration checks the saved numerical results and logical alternatives without relying on the QCA package.
- The set-fit example compares selected consistency and coverage calculations with QCA.
- The truth-table examples compare the listed rows and solutions. The LF-multiple intermediate steps also use separate mathematical checks.
- The installed app example checks existing memberships for the NF cases. It does not establish calibration or intermediate-solution agreement.
How to inspect the comparison files
Download the JSON test report to identify the software versions, settings and findings for each example. Open the CSV in a spreadsheet to inspect the available comparison values, using the result labels to locate the rows shown above.
These are reports of completed tests, not a complete rerun package. Repeating a comparison requires the original dataset and the same model, settings and software versions. The original datasets and app source code are not included.
Download the comparison results
See the values we checked, the differences we found and the software versions used. The download also identifies the original test files.
About these tests
These are QuickPLS development comparisons. Each record identifies the reference software or mathematical calculation, the tested build and the settings used.
How to understand these comparisons · Find checks for another method