QuickPLS Two plausible explanations need separate estimates
Mangundu and colleagues examined lecturers’ intentions to use cloud services in teaching. Their survey retained 202 respondents from one South African university and used PLS, including bootstrap analysis. Self-efficacy and innovativeness were among the proposed explanatory variables. The article also discusses retaining some weaker indicators rather than deleting them automatically.
In QuickPLS you will fit the four reflective constructs, check their measurement and bootstrap the two parallel indirect routes to Intention. The original teaching data let you report a clearer route alongside an uncertain one without removing either after seeing the results. They adapt the paper’s mediation question without reproducing its sample or instrument.
Your study: does perceived control relate to intention in two ways?
Imagine 300 fictional lecturers considering a cloud-based teaching service. Perceived control describes whether they feel able to manage its use. Self-efficacy concerns confidence in carrying out teaching tasks with it. Innovativeness concerns willingness to try new tools. Cloud-use intention is the outcome.
Create four reflective constructs with three indicators each. Draw Control → Self-efficacy, Control → Innovativeness, Control → Intention, Self-efficacy → Intention and Innovativeness → Intention. The two mediators are parallel: neither is drawn as causing the other in this model.
The file contains complete continuous teaching scores between 1 and 7. It is intentionally suitable for comparing a stronger and a weaker proposed route, but no particular significance decision is guaranteed. A fitted result must come from the calculation you run, not from the data-generation plan.
On a small screen, swipe the table sideways to see every column.
| Construct | Columns | Role |
|---|---|---|
| Control | control1, control2, control3 | Starting predictor |
| Self-efficacy | efficacy1, efficacy2, efficacy3 | First parallel mediator |
| Innovativeness | innovate1, innovate2, innovate3 | Second parallel mediator |
| Intention | intend1, intend2, intend3 | Cloud-use intention outcome |
| Identifier | case_id | One fictional lecturer |
Step 1. Keep capability, confidence and willingness distinct
Before analysis, explain why the three predictor-side ideas differ. Having access and control is not identical to feeling confident in a teaching task. Feeling confident is not identical to enjoying experimentation. If a real questionnaire repeats the same wording across those headings, a parallel-mediator diagram will not solve the measurement overlap.
Import the CSV through File → Import Data… → Choose data file… in a writable project. In Review & import use raw data, Header row 1, comma separation and Decimal convention Dot (1.25). Keep the twelve measurement columns Continuous. Choose Import dataset and confirm Dataset imported and verified with 300 cases.
All teaching columns point in the named direction, with no missing or reversed values. Keep case_id out of the indicators. A real lecturer survey should document invitations, responses, exclusions and context. One university’s respondents should not automatically be described as all lecturers.
Step 2. Build the five-path parallel model
Create the model
Choose File → New Model…, name it Cloud adoption pathways, confirm Dataset version and Cases, then choose Create model.
Create the four blocks
In Indicators, Ctrl-select control1, control2 and control3, then choose Create construct. Name the block Control under Properties → Model → Name and press Enter. Repeat for Self-efficacy using efficacy1, efficacy2 and efficacy3; Innovativeness using innovate1, innovate2 and innovate3; and Intention using intend1, intend2 and intend3. Set Parameter → Measurement model to Reflective for each block.
What you should see: All four constructs contain their matching indicators, with case_id left outside the model.
Draw the predictor paths
Using Connect, drag Control to Self-efficacy, Innovativeness and Intention. These three arrows include the remaining direct association as well as both first-stage paths.
Complete the parallel routes
Draw Self-efficacy to Intention and Innovativeness to Intention. Do not draw a mediator-to-mediator arrow in this planned parallel model. Return to Select and save.
Calculate the model
Choose Calculate… → PLS-SEM Algorithm. Select Starting weights → Standard (+1) and Missing data → Treatment → Casewise deletion. Review Readiness and choose Start calculation. Check the selected Result before reading the tables.
Step 3. A weaker indicator is a question to investigate
Open Estimates → Outer loadings. Read each indicator with its intended content. Then inspect Reliability and validity → Construct reliability and validity and Discriminant validity → HTMT+. A lower loading can indicate an error or a broader aspect of the concept that the other items do not capture.
Suppose a real Control item concerns access to technical help, while the other two concern personal autonomy. Removing the help item may raise a statistic while changing Control into a narrower autonomy measure. Decide whether that narrower definition answers the original question before treating deletion as an improvement.
Check coding, variation and assignment first. If the item was entered incorrectly, correct the documented error and rerun. If it is correctly entered but weak, keep the initial result and explain the content decision. Neither a published paper’s retained-item choice nor a numerical rule substitutes for that reasoning in a new study.
Hypothetical example
Illustrative reading example
Suppose Control loadings are 0.61, 0.79 and 0.83. Their mean squared loading is (0.3721 + 0.6241 + 0.6889) / 3 = 0.5617. This example shows why one modest loading and an acceptable-looking average can coexist. It does not prove content validity or instruct you to keep every weak item. These are hypothetical values.
Check the fitted answers for these 300 cases
Use all 300 cases in cloud-adoption-parallel-mediation.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. Perceived control, Self-efficacy, Innovativeness and Cloud-use intention are reflective blocks and use Mode A.
Control → efficacy is 0.514 and efficacy → intend is 0.390, giving a product of about 0.200. The innovation route is about −0.006 because innovate → intend is −0.017.
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 |
|---|---|---|
| Perceived control → Self-efficacy | 0.514 | 0.429 to 0.591 |
| Perceived control → Innovativeness | 0.332 | 0.223 to 0.436 |
| Perceived control → Cloud-use intention | 0.090 | -0.041 to 0.219 |
| Self-efficacy → Cloud-use intention | 0.390 | 0.292 to 0.490 |
| Innovativeness → Cloud-use intention | -0.017 | -0.113 to 0.093 |
How these answers were checked
The checks compared 32 estimates from the original data and 5,000 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 202611081 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
Perceived control, Self-efficacy, Innovativeness and Cloud-use intention are reflective blocks. 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 measures. Check item content, reliability and the distinction between the two mediators alongside these values.
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 |
|---|---|---|
| Perceived control / control1 | 0.953 | 0.353 |
| Perceived control / control2 | 0.947 | 0.357 |
| Perceived control / control3 | 0.938 | 0.347 |
| Self-efficacy / efficacy1 | 0.951 | 0.352 |
| Self-efficacy / efficacy2 | 0.954 | 0.359 |
| Self-efficacy / efficacy3 | 0.944 | 0.341 |
| Innovativeness / innovate1 | 0.951 | 0.344 |
| Innovativeness / innovate2 | 0.949 | 0.351 |
| Innovativeness / innovate3 | 0.943 | 0.360 |
| Cloud-use intention / intend1 | 0.943 | 0.330 |
| Cloud-use intention / intend2 | 0.952 | 0.372 |
| Cloud-use intention / intend3 | 0.947 | 0.354 |
Explained variation in each outcome
R² describes this fitted sample. It is not a test of prediction on new observations.
| Outcome | R² |
|---|---|
| Self-efficacy | 0.264 |
| Innovativeness | 0.110 |
| Cloud-use intention | 0.193 |
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 |
|---|---|---|
| Perceived control → Self-efficacy → Cloud-use intention | 0.200 | 0.137 to 0.270 |
| Perceived control → Innovativeness → Cloud-use intention | -0.005 | -0.040 to 0.031 |
Step 4. Read each indirect route on its own
Open Effects → Direct effects and read the three predictors in the Intention equation together. Self-efficacy’s path is adjusted for Innovativeness and Control; Innovativeness’s path is adjusted for Self-efficacy and Control. These are not their unadjusted correlations with intention.
Find the specific indirect Control → Self-efficacy → Intention and Control → Innovativeness → Intention outputs. Their sum is total indirect under this model. The direct Control → Intention effect is separate. Keep the route names visible in your notes so the weaker result cannot disappear inside the total.
Use Collinearity and effects → Inner VIF values to check overlapping predictors of Intention. High overlap may make the two adjusted mediator paths unstable. That is a measurement and model question, not a reason to keep only the mediator with a smaller probability.
Hypothetical example
Illustrative reading example
If Control → Self-efficacy is 0.50 and Self-efficacy → Intention is 0.45, that indirect product is 0.225. If Control → Innovativeness is 0.30 and Innovativeness → Intention is 0.06, the second product is 0.018. Total indirect is 0.243. These invented point values do not provide intervals or prove that the first route is statistically larger.
Step 5. Keep an uncertain innovation route in the output
Choose Calculate… → PLS-SEM Bootstrapping for the same saved model. Use 1000 Bootstrap samples, Confidence level (%) 95 and Seed 202611081 for the teaching run. Check the selected missing-data treatment, start calculation and read attempted, usable and failed counts.
Read each specific indirect effect with its own interval and named procedure. An interval that crosses zero means the direction or magnitude is uncertain at that level under the assumptions. It does not show that innovativeness is universally irrelevant, nor that the mediator should be removed from the planned result table.
If the self-efficacy interval excludes zero while the innovativeness interval includes it, report those two findings accurately. A claim that the routes differ needs the uncertainty of their difference. Comparing two significance labels is not that test.
Select the completed Cloud adoption pathways bootstrap and locate both specific indirect rows by their full names. Keep the direct Control → Intention row separately identified. If one interval crosses zero, retain that row and explain its uncertainty instead of replacing it with the total indirect effect.
Hypothetical example
Illustrative reading example
An innovativeness-route estimate of 0.018 with a 95% interval of [-0.025, 0.065] would be compatible with small associations in either direction. A useful sentence is: the estimated indirect association through innovativeness was small and uncertain. These bounds are hypothetical and must not be attached to a fitted estimate from your file.
Do not translate a path directly into an investment recommendation
Even a clearly positive self-efficacy association does not establish that a confidence-building workshop will increase adoption by the path coefficient. The study is about the measured variables and the specified adjustment. Intervention design, timing and other influences matter for a policy claim.
The outcome here is intention. Actual sustained use would need its own observation, period and definition. A lecturer can intend to adopt but lack institutional permission or reliable connectivity later. Keep those limits in the research question rather than burying them after a strong adoption claim.
Read the Intention R-square as fitted variation accounted for. If the practical aim is prediction for new lecturers, plan a separate prediction assessment. High explained variation and a significant indirect effect do not automatically demonstrate useful future prediction.
Decide whether the issue is measurement, overlap or uncertainty
On a small screen, swipe the table sideways to see every column.
| Observation | Check | Action |
|---|---|---|
| One Control loading is modest | Item content, coding and variation | Explain retention or revision without using a threshold alone. |
| The two mediator paths change when entered together | Their overlap and inner VIF in the Intention equation | Interpret adjusted paths and investigate measurement distinction. |
| The innovation route includes zero | Its specific product interval and plausible magnitude | Report the uncertainty; retain the planned row. |
| The total indirect looks strong | Both component routes and their signs | Do not imply that both explanations received equal support. |
| More bootstrap samples change a borderline decision | Interval endpoint stability and simulation noise | Use a declared larger plan if justified, and report fragility rather than search seeds. |
| The report says actual adoption improved | Whether the outcome is intention or observed use | Correct the outcome wording and causal scope. |
Preserve the original model when testing a refinement
A justified indicator revision should be saved separately, with its changed meaning stated. Compare measurement evidence and both indirect routes on the same cases. If a result changes substantially, explain that sensitivity rather than reporting only the version with a preferred conclusion.
Do not turn the planned parallel model into a serial one solely because one mediator is uncertain. A mediator-to-mediator arrow changes the proposed process and needs a substantive argument about ordering. If explored, identify it as a different model and retain the original parallel analysis.
Report the supported and uncertain routes together
Methods outline: We specified parallel indirect associations from Control through Self-efficacy and Innovativeness to Intention, retaining the direct Control → Intention path. All four constructs were reflective. We assessed indicator content, loadings, reliability, construct distinction and predictor overlap before examining [bootstrap procedure and settings].
Results outline: The indirect association through Self-efficacy was [estimate and interval]. The corresponding association through Innovativeness was [estimate and interval]. The direct association was [estimate and interval]. [Describe any uncertain result and measurement decision.] These findings concern reported intentions and do not establish the effect of a cloud-adoption intervention.
| Report item | Include |
|---|---|
| Parallel specification | Both mediator routes and the retained direct path. |
| Measurement decision | Weak-item evidence and the content-based reason for retention or revision. |
| Effect rows | Both specific indirect effects, their sum and the direct effect. |
| Uncertainty | Named intervals, counts, failures and any unavailable statistic. |
| Interpretation | Intention rather than actual adoption; route differences only if directly assessed. |
Put this result into a report
We fitted the specified ordinary PLS model to 300 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.
Explain the weak route without treating it as a failure
- Draw the two parallel routes without adding a mediator-to-mediator arrow.
- Recalculate the hypothetical products 0.225 and 0.018.
- Explain why a modest loading does not automatically require deletion.
- Write one sentence that retains an uncertain innovativeness result.
- Keep the planned model and any reasoned measurement revision in separate saved results.