What if the condition is not necessary? NCA using PLS-SEM in #SmartPLS4
Based on Research With Fawad's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.
SmartPLS NCA tests necessity based on antecedent coding, so absence-based necessity requires recoding the antecedent so that higher values represent absence.
Briefing
Necessary Condition Analysis (NCA) in SmartPLS-SEM is often framed around whether an antecedent’s presence is required for an outcome. This session shifts the focus: what if the outcome happens only when a condition is absent—or when different combinations of presence and absence matter? The key move is that SmartPLS-SEM’s NCA setup tests necessity tied to the coding of constructs, so researchers can probe “absence-as-necessity” by flipping scale coding so that higher scores represent the absence of the original condition.
The discussion starts by clarifying that “necessity” isn’t limited to the presence of X for Y. In theory, multiple necessity patterns exist: an antecedent’s absence can be necessary for an outcome, and necessity can also involve mixed configurations where some variables must be present while others must be absent. SmartPLS’s NCA workflow, however, is built around testing whether an antecedent construct is necessary for the presence of Y. To test absence-based necessities, the analyst must re-express the antecedent so that “absence” becomes the high-value side of the scale.
A concrete example uses leadership perceptions and organizational performance. Constructs such as role ambiguity and role conflict are measured on a 1–7 scale where low values mean low ambiguity/conflict and high values mean high ambiguity/conflict. To test whether the absence of role ambiguity and/or role conflict is associated with better organizational performance, the coding is flipped: values that originally indicated low ambiguity/conflict are transformed so that higher scores now indicate the absence of those constructs. Practically, this means transforming the dataset (e.g., in SPSS by reversing the scale for the relevant items—five items for role ambiguity and six items for role conflict) while keeping the original data intact.
After transformation, the workflow proceeds in SmartPLS 4. The transformed data are imported, the measurement model is updated to reflect the flipped constructs, and the PLS algorithm is run using unstandardized settings. Latent variable scores are then generated into a new dataset for NCA. In the NCA step, organizational performance is set as the dependent variable, while the transformed constructs act as predictors/conditions.
The necessity results are interpreted through two lenses: effect size and statistical significance. The analysis reports that role ambiguity shows a very weak necessity effect—its necessity threshold only reaches 100% of organizational performance at the extreme end—so the practical necessity is limited. A permutation test (10,000 permutations) is then used to assess whether the necessity claim is statistically supported. In this case, the permutation indicates that role ambiguity is not a necessary condition for organizational performance. The takeaway is methodological: absence-based necessity can be tested in SmartPLS NCA, but it depends on correct coding and then on verifying both effect size and significance through permutation testing.
Cornell Notes
SmartPLS-SEM NCA can test not only whether a condition’s presence is required for an outcome, but also whether a condition’s absence is required. Because the software evaluates necessity based on the coding of antecedent constructs, analysts must flip scale coding so that higher values represent the absence of the original condition. The example reverses role ambiguity and role conflict scales (originally 1–7 where high means high ambiguity/conflict) so that higher scores mean absence of ambiguity/conflict, then runs NCA with organizational performance as the dependent variable. Results are judged using effect size (e.g., necessity only reaching 100% at the extreme) and a permutation test (10,000 permutations) to check statistical significance. Here, role ambiguity’s necessity is weak and not significant.
Why can’t SmartPLS NCA directly test “absence is necessary” without extra work?
How does flipping coding translate into testing absence-based necessity?
What practical steps are needed before running NCA after recoding?
How are NCA necessity results interpreted in this example?
Why run a permutation test, and what was the outcome here?
Review Questions
- If a construct is measured so that higher scores mean “more of the condition,” what must be changed to test whether the condition’s absence is necessary for an outcome?
- What two criteria (not just one) are used to judge necessity in NCA results, and how did role ambiguity perform on each?
- In an NCA setup, which variable is treated as the dependent variable (outcome) and which are treated as predictors (conditions)?
Key Points
- 1
SmartPLS NCA tests necessity based on antecedent coding, so absence-based necessity requires recoding the antecedent so that higher values represent absence.
- 2
Necessity is not limited to “presence of X for Y”; absence of an antecedent (and mixed presence/absence configurations) can also be tested.
- 3
When flipping scales, researchers must ensure the transformed coding matches the analytical goal (e.g., higher score = absence of role ambiguity).
- 4
After recoding, the dataset must be imported into SmartPLS 4, the measurement model updated, and latent variable scores generated for NCA.
- 5
Necessity claims should be evaluated using both effect size (including how necessity reaches the outcome across the range) and permutation-based significance testing.
- 6
A permutation test with 10,000 permutations helps determine whether a weak necessity pattern is statistically meaningful; role ambiguity was not significant here.