Build your research model: Lesson 2 of PLS-SEM Course

A PLS-SEM analysis starts with a model drawn from theory. Every arrow should correspond to a hypothesis you can justify from the literature; arrows added because “they might be significant” weaken a paper. In this lesson you set up the course model in PLS-SEM Studio and learn how to translate a conceptual model into a specification.

Preparing the data

PLS-SEM Studio expects one row per respondent and one column per item, with item names in the first row. Name the items so that the construct is obvious: PE1 to PE4 for performance expectancy, BI1 to BI3 for behavioural intention. Clean the data first: remove duplicates, declare missing-value codes, and reverse-code negatively worded items (the data cleaning tool can send clean data straight to the Studio). In the Data step you choose how to handle the few remaining missing values: remove incomplete cases or replace with the item mean. Mean replacement is acceptable when less than about 5% of values per indicator are missing; otherwise, investigate why.

Try it in the tool

Open the sample study

The link opens PLS-SEM Studio and loads the sample study of 300 respondents.
In step 2 (Model), look at the six constructs and their items in the visual builder.
Switch to Text to see the same model in lavaan-style syntax.
Count the paths (6), the moderation (TR moderates PE → BI) and the four mediation routes the Studio found for you.

Open PLS-SEM Studio with the sample study ↗

Constructs and items

In the visual builder each construct is a card with its items. The course model has six reflective constructs:

Construct	Items	Example item
Performance expectancy (PE)	PE1–PE4	Using the app helps me get things done more quickly.
Effort expectancy (EE)	EE1–EE4	Learning to use the app is easy for me.
Social influence (SI)	SI1–SI3	People who are important to me think I should use the app.
Trust (TR)	TR1–TR3	I trust the app to keep my information secure.
Behavioural intention (BI)	BI1–BI3	I intend to continue using the app.
Use behaviour (UB)	UB1–UB3	I use the app frequently.

Each card has a Reflective/Formative switch. If a construct is formative, its items get outer weights instead of loadings, and the measurement model is assessed differently (weights, their significance and indicator collinearity).

Paths and hypotheses

Draw an arrow from each cause to its effect, or add paths in the Paths panel. The course model has six direct paths, which become hypotheses H1 to H6:

H1: PE → BI; H2: EE → BI; H3: SI → BI; H4: TR → BI
H5: BI → UB
H6: EE → UB (easier technologies are used more, beyond intention)

Moderation

A moderator changes the strength of a path. Here, trust is expected to strengthen the effect of performance expectancy on intention: useful apps are adopted mainly when users trust them (H7). In the Moderation panel, choose TR as the moderator and PE → BI as the path. The Studio creates the interaction term PE × TR with the two-stage approach, which Hair et al. (2022) recommend as the default.

Mediation, found for you

Whenever a construct sits on a chain of paths, there is an indirect effect. Because PE, EE, SI and TR affect BI and BI affects UB, there are four mediation routes, such as PE → BI → UB. The Studio lists them automatically and tests each one with bootstrapping. You only need hypotheses for the indirect effects that your theory cares about.

The same model in syntax

Click Text to see the model in lavaan-style syntax, which you can copy, save and edit:

PE =~ PE1 + PE2 + PE3 + PE4
EE =~ EE1 + EE2 + EE3 + EE4
SI =~ SI1 + SI2 + SI3
TR =~ TR1 + TR2 + TR3
BI =~ BI1 + BI2 + BI3
UB =~ UB1 + UB2 + UB3
BI ~ PE + EE + SI + TR + PE*TR
UB ~ BI + EE

The operator =~ defines a reflective construct (use <~ for formative), ~ defines structural paths, and PE*TR adds the interaction. Use Save project to keep the model and settings in a file you can reopen later.

Good modelling practice

Use at least three items per reflective construct when you can; single-item constructs cannot be assessed for reliability.
Keep the model as simple as the theory allows. A path you cannot justify is better left out.
Think about control variables (such as age or experience) that could explain the outcome; add them as single-item constructs with a path to the endogenous construct.
Draw the conceptual model in your paper before the results, with the hypothesis labels on the arrows.

Key takeaways

One row per respondent, one column per item, consistent item names, clean data.
Every arrow is a hypothesis justified by theory.
Moderation is added to a specific path; mediation routes follow automatically from chains of paths.
Save the model as syntax or a project file so the analysis is reproducible.