What the method does
Multi-value QCA represents conditions with several distinct categories. A category is a state, such as a type of governance, rather than a degree of membership. Configurations combine these states, and minimisation identifies simplified categorical relationships with the outcome.
Possible configurations = ∏ⱼKⱼ, where Kⱼ is the number of categories for condition j.
For two conditions with three and two categories, there are six possible combinations. Numeric category codes name states; their magnitude or distance does not imply a quantitative effect.
Data and analysis in QuickPLS
Use integer-coded categories with a clear codebook, case identifiers and an eligible outcome. Check that every observed code has the intended meaning.
QuickPLS supports the multi-value setup, categorical truth table and solution outputs. Read the category labels and condition order before interpreting a configuration.
Follow a worked example
The following example shows the data, settings and decisions for one analysis. Use the linked practice files to follow it. For your own study, keep the procedure and choose the variables, model and settings that fit your research question.
Download categories.csv to begin. It contains 6 rows and 4 columns. The example below identifies the variables to use, their roles and any rows to exclude; not every column is an analysis variable.
Keep categories as different types
Public, private and cooperative are organisation types. Codes 0, 1 and 2 can identify them, but the numbers do not make one type twice another or create a meaningful distance between types. Multi-value QCA keeps the categories distinct by mapping each original value to a named level.
Check the source text carefully. The app matches exact raw values, so spelling, letter case and spaces matter. Correct a documented coding error or add the intended mapping. Choose an outcome target that includes at least one level and leaves at least one outside, such as Achieved from an achieved/not-achieved outcome.
List the levels and the outcome you want to explain
Prepare 1 to 8 condition columns, with 2 to 5 declared levels per condition, and a case-label column. Decide which mapped outcome levels count as your target. The target must include some, but not all, outcome levels so there is a meaningful contrast.
Map the labels and calculate the categorical analysis
Open the category workflow
Select the saved dataset and use Data, More, Multi-value QCA. Choose the owning study and saved revision, then Load multi-value setup. If needed, first save a study draft in fsQCA setup; a prior fuzzy-set calculation is not required.
Map each original value
For each condition and outcome, choose Source column and Raw category type. Name the levels, enter the exact raw categories under them and write the Mapping rationale. Select the Outcome target levels.
Inspect the complete table
Set frequency, consistency/inclusion, PRI and Declared missing values. Choose Preview multi-value truth table. Check mapped cases and all condition combinations.
Save and calculate the revision
Enter a new revision name and choose Save multi-value revision. Continue in the protected copy if one is returned, then Calculate multi-value QCA.
Read one complete alternative and its cases
Inspect conservative and parsimonious alternatives separately. Read each pathway's category requirements and open its original cases to verify the source labels.
An omitted category variable allows all its levels
Inspect the complete set of condition combinations before minimization, which simplifies the eligible combinations into pathways. A pathway may require one category. If it omits a condition, it allows every declared level of that condition. Agreement for only two levels of a three-level variable is not enough to drop the variable. Check original labels beside membership in the selected categories.
A conservative solution simplifies the observed positive combinations without using unobserved ones. A parsimonious solution may use permitted unobserved combinations to simplify further. This workflow returns those two kinds of alternatives. It keeps nominal categories separate from fuzzy conditions and does not provide an intermediate multi-value solution or core/peripheral pairing. For another outcome target, save and calculate a separate specification. Plan within 1 to 8 conditions, 2 to 5 levels and the total configuration ceiling.
Level indices identify categories; a larger index does not mean more of the attribute. Read the category name in the expression. When a condition disappears from a pathway, check that the simplification really allows all of its declared categories.
The preview should show the exact sector, location and result mappings and all six configurations. Compare the mapped values with v01 to v06 in the CSV before saving. If a table requires horizontal scrolling, keep the category labels visible while reading its fit columns.
On a small screen, swipe the table sideways to see every column.
| Output | How to read it | What to check |
|---|---|---|
| Level definitions | Read each named category and its explicit mapping for every condition and outcome. | Category codes identify types; they are not degrees of membership in a fuzzy set. |
| Configuration frequency and fit | Read the observed categorical combinations under the saved thresholds. | An unobserved level combination is a remainder, not an estimated case. |
| Conservative and parsimonious alternatives | Read the category requirements in each expression and choose a complete returned alternative. | The current multi-value procedure does not offer intermediate solutions. |
| Source cases | Follow each pathway back to the original category values and any excluded cases. | Keep fuzzy-set memberships separate from nominal category inputs in this method. |
Follow the supplied category labels into a solution
Use categories.csv or a copy of the synthetic Multi-value categories project. Check the source labels against the saved mapping and results. Keep the original mapping before saving an alternative.
Study organisation type and location
The columns in categories.csv are case_id, sector, location and result. Sector is public, private or cooperative; location is urban or rural; result is achieved or not achieved. These are nominal categories: labels that distinguish types without implying numerical distances. The service-delivery meaning is invented for teaching.
The six rows contain every combination of three sectors and two locations once. Both public cases achieve the outcome. Private urban achieves it, while private rural does not. Neither cooperative case achieves it. Because the table is complete, you can check the mapping and simplification directly.
Define the mappings and the achieved target
Open Nominal categories and achieved outcome. Map public, private and cooperative to their sector levels, and urban and rural to their location levels. Set the target to achieved. Leave not achieved outside the target. Selecting every outcome level would remove the contrast you are trying to explain.
Do not assign sector codes 1, 2 and 3 and read them as fuzzy memberships or equally spaced scores. Public is not halfway between private and cooperative. If your research concept is genuinely graded, define and justify a separate fuzzy-set calibration. That is a different interpretation from this category analysis.
Build the exact mapping and inspect every combination
Load the category study
Open fsqca_multivalue.qpls, select the category dataset, then choose Data > More > Multi-value QCA. Choose the owning study and Nominal categories and achieved outcome revision, then Load multi-value setup. Check that the form identifies categories.csv. Starting from the CSV alone also requires a saved study draft; the supplied project already contains it.
What you should see: The category study and original codebook are available.
How to read it: Confirm that category values name states.
Map the two condition columns
Use sector and location as conditions. Choose Exact text for their raw categories and enter one original value per line under its named level. Keep case_id as the case label rather than an explanatory condition.
What you should see: The two condition mappings contain the intended categories.
How to read it: Numeric distances between category codes have no quantitative meaning.
Select achieved as the outcome
Choose result as the outcome source and map both achieved and not achieved. Under Outcome target levels, select achieved only.
What you should see: The outcome is achieved with its intended coding.
How to read it: Check which value represents the outcome.
Set the teaching thresholds
Use frequency 1, consistency 0.80 and PRI 0.50, with an explicit missing-data rule. These settings let the single-case configurations be inspected; they are not a universal sampling rule. Consistency measures agreement with the outcome; PRI helps identify configurations that also fit its absence.
What you should see: The frequency and fit thresholds match the teaching plan.
How to read it: Record them before examining a solution.
Preview the mappings and cases
Choose Preview multi-value truth table. Confirm all six configurations, the included cases and the readable category labels. Resolve any unmapped raw value before continuing.
What you should see: The preview shows correctly mapped cases and configurations.
How to read it: Resolve unrecognised codes before calculation.
Save and calculate the selected setup
Choose Save multi-value revision, continue in the returned protected copy when applicable, then Calculate multi-value QCA. Read an actual conservative or parsimonious complete alternative and inspect its cases.
What you should see: The completed result uses the saved mappings.
How to read it: Interpret categorical expressions through their labels.
Count the six possible combinations
Three sector levels times two location levels gives six configurations, and all are observed. The positive cases are v01 public urban, v02 public rural and v03 private urban. Cases v04 private rural, v05 cooperative urban and v06 cooperative rural are outside the achieved target.
The two public cases allow location to be omitted, giving sector = public. The private urban case remains sector = private AND location = urban. Thus the solution is Public OR (Private AND Urban), with consistency and coverage both 1 in this example. These case counts agree with the existing saved multi-value result.
On a small screen, swipe the table sideways to see every column.
| Case | Sector | Location | Target achieved |
|---|---|---|---|
| v01 | public | urban | 1 |
| v02 | public | rural | 1 |
| v03 | private | urban | 1 |
| v04 | private | rural | 0 |
| v05 | cooperative | urban | 0 |
| v06 | cooperative | rural | 0 |
Use the cooperative urban case to check a broader claim
Public covers two of the three achieved cases, so its raw and unique coverage are both 2/3. Private AND Urban covers the remaining achieved case, giving raw and unique coverage 1/3. The pathways do not overlap: a case cannot be both public and private under this single sector classification. Raw coverage counts all outcome membership captured by a pathway; unique coverage counts the part no other pathway captures. The two are equal here because the pathways are disjoint.
Urban alone would include v01, v03 and v05. Only two achieve the outcome, so its consistency is 2/3. The cooperative urban case v05 prevents the broader claim that urban location alone reproduces the perfect-fit solution. Two successful sector levels are not enough to omit a variable with three declared levels.
Investigate coding before revising thresholds
If labels differ only in spelling or spaces, check whether they refer to the same category. Correct a documented error or make an explicit mapping decision. Keep the original values and the reason for recoding. Do not merge genuinely different categories just to shorten the expression.
As a sensitivity exercise, keep the mapping fixed and raise consistency to 0.90. The completely positive single-case configurations stay positive. Raising frequency to 2 leaves no configuration with enough cases, since every combination occurs once. That exposes sparse case support; it does not justify treating category codes as ordered scores or inventing observations.
Check the size and type of the planned analysis
The supported size is 1 to 8 conditions, 2 to 5 levels per condition, up to 65,536 configurations and 10,000 original cases. Multiply the numbers of levels to count configurations. Three five-level conditions already give 125 combinations. The binary two-to-the-number-of-conditions rule does not apply when conditions have different numbers of categories.
Report the conservative or parsimonious category expressions and the actual alternatives. This workflow does not supply intermediate solutions or their core/peripheral comparison. More levels create more possible unobserved combinations, so examine the assumptions behind any remainder-based simplification as the study grows.
Resolve the mapping problem before judging the solution
Start with the original category string and the level it maps to. Then check the outcome target and the returned alternative. The table below helps distinguish a coding issue from a claim that is too broad for the observed cases.
On a small screen, swipe the table sideways to see every column.
| Symptom | What to check | Action |
|---|---|---|
| A case has no mapped level | Exact spelling, spaces and source values | Complete or correct the documented mapping before calculation. |
| The outcome has no meaningful contrast | Whether every result level was selected | Select the outcome categories you intend to explain, leaving at least one category outside the target. |
| Sector appears to have a numeric effect | Whether category codes were treated as measurements | Use the declared nominal levels and explain their meanings. |
| Urban alone seems to be sufficient | The cooperative urban case v05 | Include its contrary outcome when evaluating the broader expression. |
| Core or peripheral labels are missing | The solution types supported by multi-value QCA | Report the supported multi-value solutions without inventing those labels. |
Report the category meanings and the complete expression
Describe each condition's levels, the exact mapping and the selected outcome target. State the thresholds and case decisions. Report the complete returned alternative with its pathway and solution fit, and keep the original category labels readable.
On a small screen, swipe the table sideways to see every column.
| Pathway | Consistency | Raw coverage | Unique coverage |
|---|---|---|---|
| Public | 1 | 2/3 | 2/3 |
| Private AND Urban | 1 | 1/3 | 1/3 |
| Complete selected solution | 1 | 1 | Unique coverage applies to individual pathways |
Example Methods paragraph
We analysed six synthetic cases using sector with three nominal levels and location with two levels. The outcome target was the achieved level of result. We mapped exact source values, inspected all six possible configurations and used frequency 1, consistency 0.80 and PRI 0.50 for the teaching analysis.
Explain where each category matters
You could report: Achieved cases were covered by public organisations in either location and private organisations in urban locations. The disjoint pathways covered two-thirds and one-third of achieved cases, respectively. The complete solution had consistency and coverage of 1. Cooperative urban case v05 prevented interpreting urban location alone as sufficient.
Rebuild the six combinations from the CSV before opening the solution. Identify the case that an Urban-only expression would wrongly include, then explain why location can disappear from the Public pathway while sector remains in the Private AND Urban pathway.
- Check the declared level for every original value.
- Name the outcome category being explained.
- Multiply the condition-level counts to obtain the configuration count.
- Explain the omission of location from Public using both public cases.
- Report one actual complete alternative with its original cases.
Plan for the number of category combinations
The procedure supports up to 65,536 configurations and 10,000 cases. It treats categories as distinct levels without interpolating between their codes. Use its conservative or parsimonious results directly; fuzzy/nominal mixtures, intermediate solutions and core/peripheral pairs belong outside this workflow.
