QuickPLS What the method does
A multiple-outcome study runs distinct configurational questions within a common study design. Each outcome retains its own definition, calibration and solution. Comparable setup makes differences easier to interpret, but combining several outcomes does not turn their separate solutions into a causal chain.
For outcome j, assess X ⊆ Yⱼ using that outcome’s memberships.
The same condition expression can have different consistency and coverage for different outcomes. Preserve the outcome name and direction with each truth table and solution.
Data and analysis in QuickPLS
Provide the condition columns, each outcome column, their calibrations and a common case definition. Document any outcome-specific exclusions.
QuickPLS supports the multiple-outcome study workflow and outcome-specific results. Read each outcome separately before comparing their pathways.
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 crisp-outcomes.csv to begin. It contains 10 rows and 5 columns. The example below identifies the variables to use, their roles and any rows to exclude; not every column is an analysis variable.
Keep related research questions together
A study may ask what helps organisations adopt a service and what helps them keep using it. Those are related questions, but they need separate outcome definitions. A multiple-outcome plan keeps the analyses together while each outcome keeps its own settings and result. You decide whether they should share conditions and calibration, the rules that turn measurements into set membership.
Save each analysis setup before adding it. For outcomes Y and Z, check the meaning of each outcome, the selected conditions, calibration and cases included. The same column name can mean different things under different definitions. Share a choice only when it makes sense for both research questions.
Prepare the saved analyses you want to run
You can include 1 to 100 saved specifications, meaning saved analysis setups, from the supported enhanced, counterfactual, calibrated, multi-value and temporal families. They must refer to the same exact source dataset. Decide which conditions or calibration rules to share and which must remain specific to each outcome.
Build the plan, run it and inspect each result
Open the plan
Choose Data, More, Multiple-outcome studies. Select the Owning study that will hold the plan.
Add the saved analyses
Choose the study and saved specification for the first outcome, then click Add selected outcome. Repeat for the other outcomes and check their names and definitions.
Save the choices
Under Explicitly shared conditions and calibration, select the intended matching definition and participating items. Different saved definitions remain independent. Give the plan a useful name and click Save explicit outcome plan.
Run all planned items
Click Start / resume outcome queue. Read each item's status. A finished first item does not mean the remaining outcomes have finished.
Compare complete solutions
Click Load / refresh outcome comparison. Select an actual solution alternative for each completed item. Open the corresponding results to inspect cases and create each report.
Read the outcome name before comparing the numbers
The comparison needs a complete solution alternative that you select for each completed item. An alternative is one of the whole solutions returned by that calculation. Keep any difference in excluded cases visible. If an item failed or was interrupted, its missing result must not be read as a completed analysis that found no solution.
In this exercise Y=A and Z=1−A, so the outcomes are complements by construction. Still calculate both. Turning the Y expression into its opposite would not show what the Z calculation returned. If two items have identical complete calculation settings, QuickPLS can share their recorded calculation while keeping both item names visible.
The plan can organise enhanced, counterfactual, calibrated, multi-value and temporal analyses. It does not estimate a joint causal structure connecting all outcomes. CNA asks that different question in its own workspace; expanded mode also has a separate workflow. Prepare a report for each completed analysis you interpret.
Read the plan as a record of the work: which outcome was asked about, which cases and assumptions were used, what finished and which solution was selected. A shared calculation is a way to avoid repeating identical work. It is not a separate replication of the finding.
On a small screen, swipe the table sideways to see every column.
| Output | How to read it | What to check |
|---|---|---|
| Outcome plan item | Read each outcome, method and complete set of saved settings. | The plan runs separate analyses. It neither estimates a combined outcome model nor transfers a solution between outcomes. |
| Calculation status | Check completed, failed, cancelled and interrupted items before resuming. | Check what happened to each item; do not assume it was automatically retried. |
| Comparison of selected alternatives | Choose a complete saved alternative for each completed item before comparing them. | The table does not merge every alternative or select a winning outcome. |
| Dataset and included cases | Confirm that every item uses the same exact dataset and inspect which cases each analysis includes. | CNA, expanded fsQCA and auxiliary diagnostics use separate workflows. They cannot be added as ordinary outcome-plan items. |
Practise with two clearly defined outcomes
Use crisp-outcomes.csv or save a working copy of Multiple outcomes, CNA and crisp checks. The saved teaching results give you two outcomes to inspect and compare. Keep the original sample while building your own plan.
Study delivery and non-delivery as separate outcomes
Open crisp-outcomes.csv. Its ten invented cases have columns case_id, A, B, Y and Z. Read A as readiness, B as additional support, Y as achieved delivery and Z as non-delivery. Each membership is binary: a case is either in or out of the set. The file was deliberately made with Y = A and Z = 1 minus A so you can check the answers by counting cases.
Outcomes in a real study need not be opposites. Adoption and continued use, for example, may have different definitions and memberships. The plan helps you keep those definitions and analyses together. It does not turn them into one combined outcome or automatically decide what counts as overall success.
Check the two saved setups before adding them
Find Independent outcome Y and Independent outcome Z in the sample project. Both select A and B and include all ten complete cases. Their teaching settings are frequency 1, consistency 0.80 and PRI 0.50, with no contradiction band. Frequency is the minimum case support for a configuration; consistency measures agreement with the outcome; PRI helps identify combinations that also fit its absence. Here Y is positive for b01 to b05 and Z for b06 to b10.
Keep case_id as a label. Do not use Z as an explanatory condition for Y: in this constructed file, Z already tells you the opposite of Y. That would put the outcome definition into its own explanation. Check the reason for every selected condition before building the plan.
Build the two-item plan
After adding both items, the plan should identify Y and Z separately on the ten-case file. Read Recorded choices for each item and check both completion statuses before loading the comparison. A completed first item is not evidence that the second finished.
Open the outcome workspace
Open fsqca_multiple_outcomes_cna.qpls and select crisp-outcomes.csv. Choose Data > More > Multiple-outcome studies. Select the Owning study and locate Independent outcome Y. Confirm that the saved setup uses A and B as conditions and Y as outcome before adding it.
What you should see: The outcome workspace identifies the intended study and shared conditions.
How to read it: Check the baseline case population.
Add delivery
Click Add selected outcome. Check that the added item uses Y as its outcome, A and B as its conditions and the intended calibration and case decisions.
What you should see: Delivery is added with its definition and outcome direction.
How to read it: Confirm its calibration.
Add non-delivery
Choose Independent outcome Z and click Add selected outcome. Inspect this second item separately: adding it does not change the first item's outcome.
What you should see: Non-delivery is named and configured as intended.
How to read it: It must remain distinguishable from the delivery analysis.
Check shared settings
Under Explicitly shared conditions and calibration, choose a Condition definition and tick the intended participating outcomes. Sharing is allowed only when their saved condition and calibration choices already match. It does not overwrite different choices; outcome settings and exclusions remain separate. Read Recorded choices for each item to confirm the setup.
What you should see: Shared settings are consistent while outcome-specific choices remain visible.
How to read it: Record differences in exclusions.
Save and start
Click Save explicit outcome plan, then Start / resume outcome queue. Check the status of both named items, including any message explaining an unsuccessful attempt.
What you should see: Each requested outcome has its own calculation status and result.
How to read it: Completion for one does not establish completion for all.
Choose the results to compare
Click Load / refresh outcome comparison. Select a returned whole-solution alternative for each completed outcome and compare the expression, fit and covered cases.
What you should see: The comparison names the outcome associated with each solution.
How to read it: A separate outcome solution is not automatically a causal chain.
Keep the individual reports
Open each intended completed result and create its own report. Keep the plan alongside these reports so readers can see how the separate analyses were organised.
What you should see: Individual reports preserve each outcome’s settings and cases.
How to read it: Keep them alongside any cross-outcome summary.
Count the cases behind each expression
The A/B configurations are 00 with three cases, 01 with two, 10 with three and 11 with two. For Y, the successful configurations are those with A present; B can be absent or present. They simplify to A. For Z, configurations 00 and 01 are positive and simplify to ~A, meaning A absent.
Each outcome has five positive cases. Its selected expression covers exactly those five cases, so consistency and coverage are both 1. Consistency asks whether cases in the expression also have the outcome; coverage asks how much of the outcome the expression captures. These direct CSV counts agree with the existing saved independent results. Perfect fit follows from the file's construction.
On a small screen, swipe the table sideways to see every column.
| A B | Cases | Y-positive cases | Z-positive cases |
|---|---|---|---|
| 00 | 3 | 0 | 3 |
| 01 | 2 | 0 | 2 |
| 10 | 3 | 3 | 0 |
| 11 | 2 | 2 | 0 |
Perfect fit can describe different groups
The fit value 1 occurs in both analyses, but the solutions describe opposite case groups. Read the outcome definition and expression together. B varies within each A group, so neither selected expression needs B to describe its outcome in this teaching file.
For real outcomes, compare the included cases as well as the expressions. A missing outcome measure may remove a case from one analysis but leave it in another. Different calibration can also change memberships. A coverage difference may therefore reflect a different outcome, a different set of included cases or both; show those differences beside the comparison.
Check reasonable alternatives without changing the aim
Raise minimum frequency from 1 to 2 and all four configurations still qualify: each has 2 or 3 cases. Raise it to 3 and the two B = 1 configurations fall below the frequency cutoff. They become below-frequency rows for the analysis; this does not delete their original cases or show that B is unimportant in every setting.
A common threshold can make a study easier to organise, but each outcome still needs a defensible reason for its settings. If the reasons differ, save separate choices and explain them. Do not keep adjusting each outcome until its fit looks attractive. Use the saved setups to show what changed and why.
Follow unfinished work instead of losing it
A plan contains 1 to 100 items on one exact source dataset. It accepts supported enhanced, counterfactual, calibrated, multi-value and temporal setups; temporal items need the compatible derived temporal source. Counterfactual items retain the actual analyses for both outcome directions. CNA, panel checks, exact binomial checks, compound necessity and expanded fsQCA remain separate workflows.
Identical complete settings may share a calculation, but every named item stays in the plan. To change an existing plan, use Copy into a new editable outcome plan and save the new choices. Use Pause after current outcome when the current calculation may finish, or Cancel remaining outcomes when you want to stop the queue. Inspect failed, cancelled or interrupted items before selecting Retry selected outcomes. If the app offers Record interrupted outcome attempt after a process ended, record that state before proceeding; an interrupted item is not a completed empty solution.
Check the affected outcome's setup and status
Start with the item that looks wrong. Compare its outcome, source dataset, calibration and case exclusions with the saved setup you intended. Then read its calculation status. These checks usually tell you whether you have a different research question, a data problem or unfinished work.
On a small screen, swipe the table sideways to see every column.
| Problem | What to check | Action |
|---|---|---|
| Both items display Y | The selected specification when each item was added | Add the intended Z specification explicitly and save the corrected plan. |
| Two outcomes use unexpected shared settings | The applied shared choices and each full item setup | Retain shared choices only where intended and save a clearly revised plan. |
| An outcome comparison has a blank result | Whether that item completed successfully | Inspect its status and deliberately retry unfinished calculations where the app allows it before interpreting the comparison. |
| An unsupported specialist method cannot be added | The method of the selected specification | Use its own workflow rather than disguising it as an ordinary outcome item. |
| A comparison combines unrelated pathways | Which complete alternative was selected for each item | Select and report actual complete alternatives separately. |
| The report implies a joint solution | Whether it describes individual outcomes or the plan | Prepare individual outcome reports and keep plan-level organisation distinct. |
Make every reported result traceable to its question
State what each outcome means, which conditions and cases it used and which complete solution you selected. If settings or case counts differ, explain why. Include unfinished planned items in the account of the study instead of treating them as absent findings.
On a small screen, swipe the table sideways to see every column.
| Outcome | Positive cases | Selected expression | Consistency | Coverage |
|---|---|---|---|---|
| Y | b01 to b05 | A | 1 | 1 |
| Z | b06 to b10 | ~A | 1 | 1 |
Example Methods paragraph
We specified separate analyses for Y and Z in a ten-case synthetic dataset, using A and B as conditions. Both used existing binary memberships, all ten cases, frequency 1, consistency 0.80 and PRI 0.50. We saved an explicit two-item outcome plan and inspected a complete selected solution for each outcome.
Explain why the two answers must stay separate
A suitable Results account would say that Y was covered by A and Z by A's absence in the constructed file. Each outcome had five positive cases, with consistency and coverage of 1 for its selected expression. B was omitted from both expressions because the deliberately complementary outcome definitions were already captured by A and its absence.
Before finishing, identify the outcome and included cases of each item without relying on its position in the table. Explain to another reader why a two-item plan organises two calculations but does not estimate a two-outcome causal model.
- Read the outcome column of each added item and match it to its definition.
- Explain why the outcome's constructed complement cannot serve as an independent explanation here.
- Check every queue status and the actual complete solution selected for comparison.
- Account for outcome-specific settings and differences in included cases.
- Save the separate reports for the outcomes you interpret.
Choose the workspace that matches the question
Use the outcome plan to organise separate outcome analyses. Use the dedicated workspaces for CNA, exact binomial checks, panel diagnostics, compound necessity or expanded fsQCA. The plan does not combine those procedures into an overall causal estimate or a single combined research report.