One starting variable can have several indirect routes
Zhao and Zhao studied self-efficacy, teacher rapport, flourishing and engagement among 469 Chinese undergraduates. Their analysis used PLS with bootstrap mediation and a separate fsQCA investigation of high and low engagement.
In QuickPLS you will fit the four reflective constructs, check their measurement and bootstrap the three specific indirect routes from Confidence to Engagement. The original teaching data practise the paper’s PLS mediation analysis. Its separate fsQCA investigation asks a different question and is not reproduced by this model.
Your study: confidence, rapport, flourishing and engagement
Imagine 360 fictional students describing learning confidence, their relationship with a teacher, how well they feel they are doing in their studies, and engagement. The unit is one student. Each construct uses three reflective teaching indicators on a complete continuous 1–7 scale.
Define Confidence as the starting predictor, Rapport and Flourishing as intermediate constructs, and Engagement as the outcome. Draw Confidence → Rapport, Confidence → Flourishing, Confidence → Engagement, Rapport → Flourishing, Rapport → Engagement and Flourishing → Engagement. This six-path model contains three distinct indirect routes from Confidence to Engagement.
The route Confidence → Rapport → Flourishing → Engagement is serial. The two shorter routes go through Rapport alone and Flourishing alone. A total indirect effect adds them under this model. The file has no experimental assignment or time-separated waves, so these remain proposed statistical pathways rather than demonstrated sequences.
On a small screen, swipe the table sideways to see every column.
| Construct | Columns | Meaning in this exercise |
|---|---|---|
| Confidence | efficacy1, efficacy2, efficacy3 | Perceived ability to handle learning tasks |
| Rapport | rapport1, rapport2, rapport3 | Perceived teacher relationship |
| Flourishing | flourish1, flourish2, flourish3 | Perceived positive functioning in study |
| Engagement | engage1, engage2, engage3 | Reported involvement in learning |
| Case label | case_id | One fictional student |
Step 1. Keep the participant and the proposed time order distinct
Create a writable project and import through File → Import Data… → Choose data file…. In Review & import confirm raw data, Header row 1, a comma separator and Decimal convention Dot (1.25). Set the twelve teaching measurement columns as Continuous. Choose Import dataset and confirm 360 cases.
Keep case_id unassigned. All measurement values are complete and point in the named direction. A real engagement study would also need to explain recruitment, participation and whether students are clustered within classes or teachers. A convenient online sample does not automatically represent the wider student population.
Write down why Rapport is proposed to precede Flourishing. The diagram is your hypothesis about ordering, not evidence of it. An equally plausible account might place flourishing before perceptions of rapport, or include a common influence such as prior success. Those alternatives belong in the design and interpretation discussion.
Step 2. Create a model in which every planned route is visible
Create the model
Choose File → New Model…. Name it Student engagement routes, check Dataset version and Cases, and choose Create model.
Assign the reflective blocks
In Indicators, Ctrl-select efficacy1, efficacy2 and efficacy3, then choose Create construct. Name this block Confidence under Properties → Model. Repeat with the rapport, flourish and engage indicator sets for Rapport, Flourishing and Engagement. Under Parameter → Measurement model choose Reflective for each block.
What you should see: The four blocks contain distinct indicator sets, with case_id unassigned.
Draw the three starting paths
Choose Connect and drag Confidence to Rapport, Flourishing and Engagement. Check all arrows point from Confidence to their intended destinations.
Complete the intermediate routes
Draw Rapport to Flourishing and Engagement, then Flourishing to Engagement. Return to Select, count six arrows and save.
Calculate the point model
Choose Calculate… → PLS-SEM Algorithm. Select Starting weights → Standard (+1) and Missing data → Treatment → Casewise deletion. Review Readiness and choose Start calculation. Confirm that the completed result belongs to this six-path model.
Step 3. Make sure the mediators are distinguishable
Open Estimates → Outer loadings and read the indicators under each construct. Then inspect Reliability and validity → Construct reliability and validity. Reliability and average variance extracted help assess the reflective blocks, but they do not settle whether Rapport and Flourishing are different concepts.
Inspect Discriminant validity → HTMT+ and the actual item meanings. If flourishing items mainly repeat satisfaction with the teacher, the two intermediate constructs may overlap. A model can produce numerical indirect routes even when those routes are hard to distinguish conceptually.
Read Collinearity and effects → Inner VIF values within the Engagement equation. Confidence, Rapport and Flourishing all predict this outcome, so their separate paths are adjusted for one another. Large simple correlations do not guarantee stable adjusted paths. Check the interval widths before treating a small difference between coefficients as a useful ranking.
Check the fitted answers for these 360 cases
Use all 360 cases in engagement-paths-and-configurations.csv and the exact blocks and arrows in this lesson. The check uses ordinary PLS, path weighting, standardised indicators, starting weights +1, a 3,000-iteration limit and stopping criterion 0.0000001. Learning confidence, Teacher rapport, Study flourishing and Engagement are reflective blocks and use Mode A.
Efficacy → rapport is 0.387 and rapport → flourish is 0.380. Flourish → engage is 0.460, while rapport → engage is 0.045 after the other predictors are retained.
About these answers: Calculated from QuickPLS calculation code and independently checked with the cSEM package in R.
On a small screen, swipe the table sideways to see every column.
| Specified path | Point estimate | Reference 95% percentile interval |
|---|---|---|
| Learning confidence → Teacher rapport | 0.387 | 0.295 to 0.473 |
| Learning confidence → Study flourishing | 0.204 | 0.105 to 0.309 |
| Learning confidence → Engagement | 0.245 | 0.157 to 0.331 |
| Teacher rapport → Study flourishing | 0.380 | 0.283 to 0.465 |
| Teacher rapport → Engagement | 0.045 | -0.052 to 0.139 |
| Study flourishing → Engagement | 0.460 | 0.377 to 0.542 |
How these answers were checked
The checks compared 33 estimates from the original data and 6,000 direct-path estimates from 1,000 repeated fits. Every difference was within the absolute tolerance of 0.00000001. QuickPLS v1.03.11 calculation code and cSEM 0.7.1 used the same model, file and sampled rows. Both calculations returned usable results for all 1,000 fits.
The reference used R seed 202611079 to sample rows and calculated percentile intervals with R quantile type 7. It repeatedly fitted the model through calculation code; it did not use the desktop app’s Bootstrapping command. The same seed can produce different sampled rows in different programs. The download includes the actual row numbers, unrounded answers and settings so this calculation can be repeated.
Check the measurement values and explained variation
Check the measurement values and explained variation
All four blocks in this exercise are reflective. A loading relates an indicator to its construct score; the weights are used to form those scores. The weight column does not make these formative blocks. Read the loadings alongside reliability, discriminant-validity and item-content checks; high loadings alone do not establish validity or distinct concepts.
About these answers: Estimates from the original data, calculated with QuickPLS code and cSEM using the same model and file.
On a small screen, swipe the table sideways to see every column.
| Block / indicator | Loading | Weight |
|---|---|---|
| Learning confidence / efficacy1 | 0.939 | 0.349 |
| Learning confidence / efficacy2 | 0.942 | 0.347 |
| Learning confidence / efficacy3 | 0.947 | 0.365 |
| Teacher rapport / rapport1 | 0.944 | 0.340 |
| Teacher rapport / rapport2 | 0.951 | 0.352 |
| Teacher rapport / rapport3 | 0.951 | 0.362 |
| Study flourishing / flourish1 | 0.959 | 0.357 |
| Study flourishing / flourish2 | 0.948 | 0.346 |
| Study flourishing / flourish3 | 0.953 | 0.346 |
| Engagement / engage1 | 0.955 | 0.364 |
| Engagement / engage2 | 0.941 | 0.344 |
| Engagement / engage3 | 0.956 | 0.344 |
Explained variation in each outcome
R² describes this fitted sample. It is not a test of prediction on new observations.
| Outcome | R² |
|---|---|
| Teacher rapport | 0.150 |
| Study flourishing | 0.246 |
| Engagement | 0.381 |
Read each indirect route separately
Each row follows one specific route. Multiply its component coefficients within each repeated fit before taking the interval; multiplying the endpoints of component intervals is incorrect. These are statistical associations in synthetic data, not evidence of causal mediation.
About these answers: Each indirect effect was recalculated in every saved reference resample. These intervals were not produced by the desktop app’s Bootstrapping command.
On a small screen, swipe the table sideways to see every column.
| Specific route | Point product | Reference 95% percentile interval |
|---|---|---|
| Learning confidence → Teacher rapport → Study flourishing | 0.147 | 0.097 to 0.201 |
| Learning confidence → Teacher rapport → Study flourishing → Engagement | 0.068 | 0.043 to 0.098 |
| Learning confidence → Teacher rapport → Engagement | 0.017 | -0.020 to 0.057 |
| Learning confidence → Study flourishing → Engagement | 0.094 | 0.048 to 0.146 |
| Teacher rapport → Study flourishing → Engagement | 0.175 | 0.121 to 0.230 |
Step 4. Read specific indirect effects before their sum
In Effects locate the specific indirect outputs. Write each route in full beside its estimate: Confidence → Rapport → Engagement; Confidence → Flourishing → Engagement; and Confidence → Rapport → Flourishing → Engagement. Do not shorten all three labels to mediation, because they represent different products.
The direct Confidence → Engagement path is separate. Total indirect adds the three specific routes, while total effect combines direct and indirect parts under the specified linear model. Keeping the components visible helps you explain whether the total is dominated by one route or spreads across several.
Use full-precision coefficients from this same result for a manual check. If the handwritten product differs substantially, compare the named path, scale, cases and model version before blaming rounding. A coefficient from a model that omitted Flourishing is not interchangeable with the Rapport coefficient in this fuller model.
Hypothetical example
Illustrative reading example
Suppose Confidence → Rapport is 0.40, Rapport → Engagement is 0.20, Confidence → Flourishing is 0.30, Rapport → Flourishing is 0.40 and Flourishing → Engagement is 0.50. The short Rapport route is 0.08, the short Flourishing route is 0.15, and the serial route is 0.08. Total indirect is 0.31. These invented coefficients teach the decomposition; they are not answers from the CSV.
Step 5. Do not rank routes from separate significance decisions
Choose Calculate… → PLS-SEM Bootstrapping on the saved six-path model. For this teaching run set Bootstrap samples to 1000, Confidence level (%) to 95 and Seed to 202611079. Read the actual interval method and attempted, usable and failed counts after calculation.
Give every specific indirect estimate its own interval. The complete-model bootstrap recalculates the products together in each resample. It does not follow that one route is stronger merely because its interval excludes zero while another route’s interval includes zero.
If your research question asks whether two indirect routes differ, that difference needs its own uncertainty assessment. Do not invent a difference test by subtracting p-values or comparing which row has an asterisk. This lesson supplies route-specific inference; report a numerical ranking as descriptive unless an appropriate direct comparison is actually performed.
Select Student engagement routes in the completed bootstrap result. Locate Confidence → Rapport → Engagement, Confidence → Flourishing → Engagement and Confidence → Rapport → Flourishing → Engagement by their full labels. Read each interval beside its own estimate before examining total indirect.
Hypothetical example
Illustrative reading example
Imagine route A is 0.08 with interval [-0.01, 0.17], while route B is 0.15 with interval [0.04, 0.27]. B is more clearly positive under these separate intervals, but this does not establish that B minus A differs from zero. The two estimates share cases and model parameters. These are hypothetical intervals.
Why fsQCA would answer a different engagement question
The source paper also used fsQCA, which studies combinations of conditions associated with an outcome. A PLS indirect route describes a product of adjusted relationships. An fsQCA pathway describes a set-membership configuration. They are not two interchangeable tests of the same mediation coefficient.
For example, a future configurational question might ask whether high rapport and high flourishing together accompany high engagement. That question would require explicit set definitions and calibration, then truth-table and case analysis. A positive PLS path does not supply a calibration anchor or prove that the condition is necessary.
Low engagement would need its own outcome-absence analysis. It is not automatically the mirror image of the explanation for high engagement. The linked fsQCA lesson walks through those choices with a separate original file; do not attach its solution to the PLS results from this exercise.
Keep the comparison matched to the question
On a small screen, swipe the table sideways to see every column.
| What happens | Check | What to do |
|---|---|---|
| Two indirect rows look identical | The complete route names and component coefficients | Keep short and serial routes separate even when their estimates coincide. |
| One route is significant and another is not | Whether a direct contrast was calculated | Do not claim the routes differ from separate decisions alone. |
| Rapport’s adjusted path changes sign | Overlap, coding and other predictors of Engagement | Inspect the full equation and uncertainty before telling a causal story. |
| Total indirect hides a negative component | Signs of all specific routes | Report the components and explain cancellation. |
| The product interval is unavailable | Failed refits and statistic-specific messages | Report missing inference accurately rather than borrowing a path interval. |
| A PLS route is called a necessary condition | Whether a set-based necessity analysis exists | Use association language and keep the two methods’ questions separate. |
Make an ordering comparison explicit
If theory genuinely supports an alternative ordering, save it as a separate exploratory model and explain the changed question. Do not fit both orders, choose the more significant one and describe that order as established. Cross-sectional relationships alone usually leave important ordering questions unresolved.
A practical measurement sensitivity check can examine whether the conclusion changes under a justified refinement of Rapport or Flourishing. Keep the original items and model on record. Describe the content lost or clarified by the revision, then compare the same named routes on the same cases.
Name the route in every result sentence
Methods outline: We fitted a reflective four-construct PLS model with the six stated paths. We assessed measurement quality, construct distinction and predictor overlap before interpreting the three specific indirect associations from Confidence to Engagement. We obtained [interval method] intervals from [bootstrap settings], with [usable and failed counts].
Results outline: The indirect association through Rapport alone was [estimate and interval], through Flourishing alone [estimate and interval], and through Rapport then Flourishing [estimate and interval]. The total indirect association was [value]. These route-specific results do not establish that one indirect association is larger than another without a direct comparison, or that the proposed order is causal.
| Report row | Required label |
|---|---|
| Short route 1 | Confidence → Rapport → Engagement |
| Short route 2 | Confidence → Flourishing → Engagement |
| Serial route | Confidence → Rapport → Flourishing → Engagement |
| Direct and total | Their own definitions, estimates and intervals |
| Inference | Exact procedure, usable counts and any unavailable statistic |
| Design | Measurement timing and plausible alternative orderings |
Put this result into a report
We fitted the specified ordinary PLS model to 360 original synthetic teaching cases. The measurement roles and complete structural model were fixed before interpreting the results. Point estimates were checked against an independent cSEM calculation.
Report the relevant path estimate and its matching interval from the table. If you use these supplied intervals, identify the separate repeated-fit reference calculation. For results from your own app analysis, report that run’s settings and intervals. A real study also needs the actual sampling design, measurement evidence and a reason for each model choice.
Check the decomposition and the comparison
- Identify the three indirect routes without looking at the output labels.
- Use the saved full-precision coefficients to calculate the three Confidence-to-Engagement indirect products and their sum. Compare their separate reference intervals without treating significance labels as a direct difference test.
- Explain why a significant route and a nonsignificant route are not automatically different.
- Describe how a configurational question differs from an indirect path.
- Save the complete planned model and label any ordering revision as exploratory.