# 26. Moderation

## Purpose and applicability

Moderation asks whether the magnitude or direction of a relationship changes across values of another construct. This tutorial examines whether Organizational Prestige changes the relationship between Organizational Identification and two forms of Affective Commitment.

Use moderation when theory specifies a focal relationship, a moderator, and an outcome before estimation. Do not interpret an interaction as causation, and do not omit the focal and moderator main effects from the model.

## Practice study

- Dataset: [organizational-identification-model-comparison.csv](../samples/organizational-identification-model-comparison.csv)
- Prepared project: [organizational-identification-moderation.qpls](../projects/organizational-identification-moderation.qpls)
- Observations: 305 complete cases
- Variables: 22
- Data type: deterministic packaged teaching study
- Research question: Does Organizational Prestige change how Organizational Identification relates to Affective Commitment (Joy) and Affective Commitment (Love)?

This is the complete 305-case study packaged with QuickPLS. No cases were invented or duplicated to meet the tutorial catalogue's usual sample-size range.

## Model and variable roles

The prepared model contains four reflective constructs:

- Organizational Prestige: `org_pre1`–`org_pre8`
- Organizational Identification: `org_ident1`–`org_ident6`
- Affective Commitment (Joy): `ac_joy1`–`ac_joy4`
- Affective Commitment (Love): `ac_love1`–`ac_love3`

The saved structural model estimates Organizational Prestige → Organizational Identification; Organizational Prestige and Organizational Identification → Affective Commitment (Joy); and Organizational Prestige and Organizational Identification → Affective Commitment (Love).

It also contains two authored two-way interaction terms:

- Organizational Identification × Organizational Prestige → Affective Commitment (Joy)
- Organizational Identification × Organizational Prestige → Affective Commitment (Love)

The `gender` column is retained as a descriptive field and is not used in this moderation calculation.

## Exact procedure

1. Start QuickPLS, choose **Open Project**, and select [organizational-identification-moderation.qpls](../projects/organizational-identification-moderation.qpls).

2. Open **Data**. Confirm 305 observations and 22 variables. The linked CSV is available separately for inspection and re-import practice.

   ![Organizational-identification data prepared for two-outcome moderation](../screenshots/moderation/01-data-ready.png)

3. Return to **Model** and confirm the four reflective measurement blocks, the five main-effect paths, and both interaction terms listed above.

4. The prepared project stores the qualified coordinates used by this tutorial. It opens already arranged and fitted; do not press **Arrange** or move model objects before calculation.

   ![Saved arranged two-outcome moderation model](../screenshots/moderation/02-arranged-model.png)

5. Choose **Validate**. The model must have no blocking issue. If you intentionally edit the model, create and save a project revision before calculating.

6. Use **Calculate → PLS-SEM Algorithm**. QuickPLS detects the two authored moderation relationships and lists both in the calculation setup.

   ![PLS-SEM Algorithm setup showing both authored moderation relationships](../screenshots/moderation/03-calculation-setup.png)

7. Confirm that **Open Results when finished** is selected and choose **Start calculation** once. Wait until the point-estimation result is saved and Results opens.

8. In **Results**, select the completed **General SEM simultaneous two-way PLS moderation point estimates** result. Inspect these outputs:

   1. **Interaction effects and conditional slopes**
   2. **Conditional focal slopes**
   3. **Interaction plot points**
   4. Both moderating-effect charts

9. Open **Conditional focal slopes**. The ordinary table presents the moderating effect, moderator, standardized moderator value, outcome, and conditional slope; exact technical identities remain available through **Columns**.

   ![Conditional focal slopes for both affective-commitment outcomes](../screenshots/moderation/04-results-primary.png)

10. Open the first **Moderating effect** chart. Compare the three lines for standardized Organizational Prestige values of −1, 0, and +1, and use **Exact plot data: Interaction plot points** when exact values are required.

    ![Interaction chart for Affective Commitment Joy](../screenshots/moderation/05-results-secondary.png)

11. Choose **Save Report** to preserve the selected result, or use **Copy** and **Export** for the needed table format. Close and reopen the prepared project once and confirm that the saved result remains selectable.

## Reading the saved point result

For Affective Commitment (Joy), the conditional slope of Organizational Identification rises from 0.4989 at Prestige = −1 to 0.5494 at the mean and 0.5998 at Prestige = +1. For Affective Commitment (Love), the corresponding point estimates are −0.2055, −0.4043, and −0.6031. The chart visualizes these fitted point estimates; it does not by itself establish statistical significance.

The saved interaction coefficients are approximately +0.0505 for Joy and −0.1988 for Love. Interpret their direction together with the conditional slopes and the research design, not as stand-alone causal findings.

## Bootstrap inference

This qualified prepared result contains simultaneous moderation point estimates. It does **not** contain bootstrap standard errors, confidence intervals, or p values. If inference is required, first confirm this point model and then run the applicable PLS full-model bootstrap workflow. Report the chosen bootstrap settings and interpret an interaction inferentially only when its saved interval and p value are available.

Do not report a Johnson–Neyman region from this tutorial: the qualified 2.62.8 point result provides low, mean, and high moderator probes, not a Johnson–Neyman output.

## Common mistakes and limitations

- Do not create a second interaction term when the prepared project already contains it.
- Do not treat the `gender` field as the moderator; the moderator here is Organizational Prestige.
- Do not infer significance from line separation or a point coefficient alone.
- Do not press **Arrange** in the prepared project; its presentation coordinates are already saved.
- The data are deterministic teaching data. They reproduce the workflow but are not substantive empirical evidence.

## Related methods

- [Mediation](27-mediation.md) evaluates indirect effects rather than interaction effects.
- [PLS-SEM Bootstrapping](02-pls-sem-bootstrapping.md) explains the ordinary PLS bootstrap result families.
- [PLS-SEM Conditional Process](28-conditional-process.md) combines mediation and moderation in one model.

Last verified with QuickPLS 2.62.8.