QuickPLS The research behind this exercise
Kim and colleagues studied 444 residents of Sacheon and Jinju using PLS-SEM, multigroup analysis and fsQCA. Their model considered development principles, environmental and social governance perceptions, support and quality of life. The reported PLS-MGA comparisons did not establish path differences at the conventional .05 level. This lesson teaches the direct contrast using original residential groups, rather than the paper’s support-based split.
In QuickPLS you will fit the common resident model, review MICOM and compare the planned Fairness → Support path between GROUP A and B. This original exercise uses a predefined residence distinction to practise a direct group comparison. It does not reproduce the paper’s support-based split or its fsQCA analysis.
The question you will answer
Imagine a destination considering a visitor attraction beside a new transport facility. Residents living nearby and farther away may have different experiences. Is the relationship between perceived fairness and support actually different between those groups?
The GROUP column defines A as the fictional nearby-resident group and B as the fictional farther-away group. These labels are assigned for teaching; the file does not promise different path strengths. Keep GROUP outside all indicator blocks. Use the same model and indicator meanings for both groups.
Define group membership before looking at support scores. Splitting people into high and low support would condition the comparison on a model outcome and answer a different question. The original exercise uses a pre-existing residence distinction so the group definition is easier to explain.
Step 1. Download the original practice file
The download has 400 complete fictional records: one fictional resident per row. Its continuous values sit within a 1–7 teaching response range; they are not rounded survey categories. The number of rows makes the exercise convenient to practise. It is not a sample-size recommendation for a real study.
Open the codebook before importing. Check the meanings and direction of the columns, then confirm the row count and numeric values in QuickPLS. Keep case_id only as a row identifier. It is not an indicator, predictor or control. These files have no planned missing entries, so an unexpected blank calls for an import check.
For your own study, distinguish a missing response from a genuine zero and document exclusions before analysis. Reverse a question only when its wording and scoring require it. The practice columns are already oriented as described below.
After import, check the FAIR, ENV, SUP and QOL measurement columns and the separate GROUP labels. Keep GROUP outside the indicator blocks and preserve its predefined A/B meaning.
Open the import
Open a writable project and choose File → Import Data… or Ctrl+I. In Import dataset click Choose data file… and select this lesson’s CSV.
Review the preview
The dialog moves from Choose file to Review & import. Set Input represents to raw data, Header row to 1, separator to comma and Decimal convention to Dot (1.25). Confirm the teaching columns are Continuous. Keep case_id out of the analysis.
Complete the import
Check that the preview separates the columns correctly, then click Import dataset. Wait for Dataset imported and verified. Open the imported data and confirm the expected case count before building the analysis.
On a small screen, swipe the table sideways to see every column.
| Idea or role | Columns | How this exercise uses them |
|---|---|---|
| Perceived fairness | FAIR1, FAIR2, FAIR3 | reflective construct |
| Environmental confidence | ENV1, ENV2, ENV3 | reflective construct |
| Development support | SUP1, SUP2, SUP3 | reflective construct |
| Expected quality of life | QOL1, QOL2, QOL3 | reflective construct |
| A: fictional nearby residents; B: fictional farther-away residents | GROUP | Group label, kept outside the measurement blocks |
Step 2. Build the measurement blocks and arrows
A construct is an idea represented by several measured columns. An indicator is one of those columns. Here each three-column block is reflective: the three scores are treated as expressions of the same idea. This is a teaching assumption, not evidence that a newly written questionnaire is valid.
Use the exact assignments in the column table. The planned structural paths are Perceived fairness → Development support; Environmental confidence → Development support; Development support → Expected quality of life.
Create the model
With the imported data open, choose File → New Model…. Give the model a distinct name, select this lesson’s file under Dataset version, check Cases and click Create model.
Create the first block
In Indicators, Ctrl-select FAIR1, FAIR2 and FAIR3. Check 3 selected and click Create construct.
What you should see: The first block contains the Perceived fairness indicators; GROUP remains unassigned.
Name it and set measurement
Choose Select and click the construct body. If needed, open View → Show Properties. Under Properties → Model enter the name from the column table and press Enter. Under Parameter → Measurement model choose Reflective.
Add the remaining blocks
Repeat that selection, naming and measurement check for every three-column block. Keep each column in its intended block. To fix an assignment, use Properties → Data Binding → Remove, then select the right column and use Assign to [construct name].
Draw and save the planned paths
Choose Connect. Drag from the body of each source construct to the body of its destination, then release. Follow the arrows listed above and return to Select. Check arrow direction and save the model. An arrow records a proposed relationship; it does not establish cause and effect.
Step 3. Run the model and find the measurement checks
PLS-SEM forms scores from the indicators and estimates the relationships in your diagram. First establish that the intended file and model can be fitted. Resampling and extensions come afterwards.
Before the group analysis, confirm that the pooled model contains FAIR, ENV, SUP and QOL with the stated paths. The same measurement and structural specification will be used in both groups.
Open the calculation
Click Calculate… and enter PLS-SEM Algorithm in Find a method. Open Methods if the method list is collapsed, then select the matching row.
Check the visible choices
Use Starting weights → Standard (+1) and Missing data → Treatment → Casewise deletion. The latter keeps complete cases; the supplied file is complete. Review Readiness, leave Open Results when finished selected and click Start calculation.
Select the completed run
In Results, select the new calculation in the Result dropdown. Confirm the analysed cases and convergence. A calculation that did not settle to a stable answer needs attention before its coefficients can be interpreted.
Open the measurement tables
Open Estimates → Outer loadings, Reliability and validity → Construct reliability and validity, and Discriminant validity → HTMT+. Search result outputs can locate a table; clear the search afterwards to restore the full folder list.
Step 4. Check the questions before explaining the arrows
Ask whether Fairness means the same thing for both residential groups. An item about access to benefits might be interpreted differently by people experiencing disruption near the site. The same column names and good pooled reliability do not establish measurement comparability. Review the common specification and then use MICOM before interpreting the affected path contrast.
Read a loading with the meaning and direction of its item. Reliability concerns how consistently a block works together. Average variance extracted, or AVE, summarises how much indicator variation the reflective block represents. HTMT+ compares pairs of reflective constructs and helps you investigate whether supposedly different ideas are too alike.
Open Collinearity and effects → Inner VIF values for predictors entering the same outcome. Strong overlap makes their separate contributions harder to distinguish. A threshold is a prompt to investigate, not a replacement for reading the item definitions. Record concerns before inspecting which hypothesis looks attractive.
Inspect the Fairness and Support measures in this result. Their item definitions and the group-comparability checks matter for the selected FAIR → SUP contrast; pooled reliability alone does not establish comparability.
Add intervals for the planned relationships
Return to Calculate… and choose PLS-SEM Bootstrapping. For this repeatable practice run set Bootstrap samples to 1000, Confidence level (%) to 95 and Seed to 261107111. These are teaching choices, not a claim that the paper used them or that 1000 draws is always enough for final research.
Start the calculation and inspect requested, attempted, usable and failed draws. A bootstrap repeatedly draws complete rows with replacement and refits the model. Read the interval name and test definition supplied with the result; the basic complete-case panel does not provide the separate PLSc interval selectors.
Open Effects → Direct effects for the drawn arrows. For a pathway through an intermediate construct, find its named specific indirect effect and use that row’s interval. Open Estimates → R-squared for explained variation and Collinearity and effects → f-square effect sizes for each predictor’s contribution. Keep the original estimate, interval and output label together.
Keep this ordinary PLS bootstrap separate from the group procedure below. The selected FAIR → SUP difference needs the group workflow’s own permutation result, not a comparison of two unrelated significance labels.
Step 5. Set up the direct group comparison
Open the combined workflow
Choose Calculate… → MICOM and Multigroup Analysis (PLS / PLSc) → Continue to setup. Set Grouping column to GROUP and select A and B. Check their complete-case counts.
Review the common measurement
Choose Profile envelope → Recursive PLS-SEM. Complete the MICOM Step 1 checklist only after reviewing common indicators, coding, missing-data treatment, settings, model specification and sign orientation.
Choose the procedures and parameter
Select MICOM Steps 2–3, pairwise and Pairwise size-preserving permutation. Under Parameters to compare select Perceived fairness → Development support. This is the single planned contrast.
Record the rule and run
Set Permutation and bootstrap draws to 5000, Alternative to Two-sided and Multiplicity to Holm FWER. Record the seed and worker settings. Click Validate setup, resolve blockers and then Start calculation.
Check comparability before the group difference
Use GROUP as a text grouping column, with A first and B second. Keep it outside the measurement blocks. The file contains 200 A cases and 200 B cases. Fit the same four reflective blocks and the same paths in both groups.
Select MICOM and the pairwise permutation comparison, choose only FAIR → SUP, set two-sided inference, Holm adjustment, 5,000 permutations, seed 202611111 and one worker. Review the six Step 1 statements about identical coding, data treatment, model, settings, sign orientation and analyst review before accepting comparability.
All four constructs meet the Step 2 composition criterion in this reference.
The FAIR → SUP estimate is 0.491 in A and 0.513 in B. The A-minus-B difference is -0.022; the two-sided reference permutation probability is 0.770. This teaching comparison does not provide clear evidence that the path differs between the groups.
About these answers: Group comparisons calculated independently in cSEM and R. Estimates from the original groups were also checked with QuickPLS calculation code; the full app group-comparison procedure was not run.
On a small screen, swipe the table sideways to see every column.
| Construct | Composition correlation | Lower 5% reference boundary | Step 2 criterion met |
|---|---|---|---|
| FAIR | 0.999947 | 0.999616 | Yes |
| ENV | 0.999647 | 0.999291 | Yes |
| SUP | 0.999998 | 0.999850 | Yes |
| QOL | 0.999942 | 0.999766 | Yes |
What was repeated and what was checked
5,000 permutations requested; 5,000 usable; 0 failed. Each saved assignment keeps the two group sizes fixed. Both group models are refitted and their scores are aligned to one pooled reference. The same assignments supply MICOM and the selected path contrast.
The reference uses R’s random generator and saves every reassignment of cases to groups. The app uses a different generator, so entering the same seed does not guarantee identical permutation probabilities. The estimates for the original groups were checked separately; the full app permutation procedure was not compared.
QuickPLS ordinary PLS calculations and NumPy checked the original-sample results for the pooled data and separate groups: score composition, means, log-variance and the selected paths. All Step-3 permutation statistics, tail probabilities and quantiles were checked with the pooled scores held fixed. The cSEM group models fitted after each permutation were not independently rerun. These answers therefore come from the separate R reference, not the app’s MICOM/MGA calculation.
The configuration may require a bootstrap count, but no bootstrap procedure is selected here. These are 5,000 permutations, not 5,000 bootstrap samples. One selected path makes Holm’s adjusted path probability equal to its raw probability.
Keep composition, means and variances separate
No construct mean difference is detected. The variance check detects a difference for QOL.
Step 3 uses the fixed pooled construct scores. Its variance statistic is log(variance in A) minus log(variance in B), with sample variances using n−1. A positive value means greater variance in A. A non-significant check means that this procedure did not detect a difference; it does not prove equality.
On a small screen, swipe the table sideways to see every column.
| Construct | Mean A−B | Mean p | Log-variance A−B | Variance p |
|---|---|---|---|---|
| FAIR | -0.072 | 0.477 | 0.092 | 0.505 |
| ENV | 0.120 | 0.232 | -0.044 | 0.736 |
| SUP | -0.013 | 0.886 | -0.022 | 0.867 |
| QOL | 0.072 | 0.470 | 0.303 | 0.032 |
Read the one planned path contrast
The FAIR → SUP estimate is 0.491 in A and 0.513 in B. The A-minus-B difference is -0.022; the two-sided reference permutation probability is 0.770. This teaching comparison does not provide clear evidence that the path differs between the groups.
The source and target constructs both meet Step 2, so the selected path passes this reference’s composition check once the reviewed Step 1 conditions are accepted.
A mean difference, a variance difference and a path difference answer different questions. Do not use one as a substitute for another, and do not infer a group difference merely because one within-group path is significant and the other is not.
On a small screen, swipe the table sideways to see every column.
| Path | A | B | A−B | Two-sided permutation p |
|---|---|---|---|---|
| FAIR → SUP | 0.491 | 0.513 | -0.022 | 0.770 |
Step 6. Read comparability before the difference
Start with Group eligibility and retained cases. Read the MICOM steps for Fairness and Support and any other affected construct. Step 1 concerns a common specification. Step 2 concerns comparable component formation. Support for both provides partial invariance for the relevant standardized comparisons. Step 3 separately examines means and variances.
Next inspect the group estimates and the A-minus-B contrast for Fairness → Support. Use that contrast’s own permutation inference. Keep its group order and comparison family visible. Do not combine an A estimate from one saved model with a B estimate from a later revision.
Read Group eligibility and the MICOM decisions for Fairness and Support in this resident result. Then open their A-minus-B path comparison and retain the group order, test and eligibility information with the estimates.
One significant path and one uncertain path can still be similar
An estimate and its precision both affect its individual test. Different precision across groups can produce different significance labels even when the coefficients themselves are not detectably different.
Hypothetical example
Illustrative reading example
Suppose the group A coefficient is 0.40 and group B is 0.28. Their A-minus-B difference is 0.12. Even if one within-group path is significant and the other is not, a direct contrast p-value of 0.31 would not support a group difference under that test. These are invented values, not predictions for the teaching groups.
What to check when the result is unexpected
If MICOM Step 2 fails for an affected construct, investigate coding, item interpretation and sample composition before interpreting its group path contrast. A coefficient difference cannot repair a measurement-comparability problem.
If one group is much smaller after exclusions, inspect why. Complete-case rules can change the composition of the comparison. Do not merge unrelated residence categories solely to satisfy a software size check.
If the contrast is uncertain, report its size and uncertainty. A nonsignificant difference does not prove the groups are identical. It means this comparison did not provide sufficient evidence of a difference under the specified analysis.
Save the original model and the reason for each genuine correction. Keep unsupported relationships in the results you report. A more convincing study explains a weak or uncertain finding and plans better evidence; it does not erase cases, items or paths until every test is favourable.
Write a clear Methods and Results account
Methods outline: We compared the predefined groups [definitions and counts] under the same reflective PLS specification. After the MICOM review, we tested the planned Fairness → Support difference using [permutation settings and adjustment].
Results outline: The relevant MICOM findings were [stage-specific results]. The coefficients were [A] and [B], with A-minus-B difference [value] and [contrast inference]. We therefore concluded [bounded group statement], rather than comparing the separate path significance labels.
Include the analysed case count, exact file or data version, model diagram, measurement decisions, calculation settings and the uncertainty procedure. Attach the relevant full tables rather than only the strongest row. Explain which decisions were planned and which followed inspection.
For this exercise, call the records simulated teaching data. In an actual paper, replace every bracketed placeholder with the result from the named saved run. Remove any illustrative numbers from this lesson before drafting your findings. Cite the original study for the research context and the relevant methodological references for the analysis.
Report the comparison you actually tested
State the group definitions, counts, reviewed configuration checks, composition results and the selected A-minus-B path. Keep Step 3 location and spread results separate. Report the requested and usable permutation counts, the alternative and the adjustment family.
For these reference answers, identify the independent cSEM/R calculation and its saved assignments. For your own app analysis, report the settings and results exported from that actual run.
Check your conclusion before you finish
- Define groups independently of the outcome being compared.
- Report relevant MICOM stages before the path contrast.
- Use the difference’s own test.
- State group order, retained counts and adjustment family.