nomologR Vision Charter • nomologR

Why this package exists

Scale development is often taught as a sequence of disconnected statistical procedures: inspect alpha, run an EFA, delete weak items, run a CFA, report fit, and declare the measure valid.

That workflow is easy to execute badly because the difficult part is not running the analyses. The difficult part is understanding why an analysis is appropriate, what evidence it contributes, how one decision changes the next, and when a result should alter the theory rather than merely the model.

nomologR exists to make that reasoning visible.

Who nomologR serves

Graduate students (master’s and doctoral) learning scale development and construct validation, who need to understand why each analysis is performed.
Faculty and advisors who teach measurement and want output that explains its own reasoning.
Researchers applying these techniques who need a defensible, reproducible record of how a measure was evaluated.

The package should be usable by the first group and trustworthy to the last.

What nomologR is

A guided layer around established psychometric and SEM tools that:

chooses or recommends methods appropriate to the data,
explains the methodological rationale,
produces transparent diagnostics,
records researcher decisions,
connects measurement evidence to theory,
and teaches while it analyzes.

What nomologR is not

not a replacement for psych
not a replacement for lavaan
not a replacement for semTools
not an automatic item-deletion algorithm
not a score that declares a scale “valid”
not a black-box SEM generator
not a content-validity package
not a Solomon-design package

The defining question

At every stage, the package should help answer:

What does this result tell me about the claim that these observations measure the construct I say they measure, and what should I investigate next?

The core intellectual path

Concept → items → content evidence → dimensionality → measurement model → reliability → construct-validity evidence → invariance/generalizability → theoretically predicted network → continued accumulation of evidence.

contentvalidR primarily supports the front of this path. nomologR primarily supports the empirical measurement and network portion. solomonR addresses a separate experimental-design problem.

Standard for recommendations

A recommendation in nomologR should have four parts:

Observation — what was found.
Reason — why it might matter.
Options — defensible choices available to the researcher.
Consequence — what each choice means for subsequent analysis.

Example:

Item S4 has a primary loading of .37, below the teaching reference of .40. Its communality is .51 and it has no material cross-loading. This is weak evidence against the item, not an automatic deletion criterion. If the item represents unique theoretical content, retaining it may preserve construct breadth. If similar content is already represented by stronger items, removal may improve measurement precision. Compare both models and record the rationale for the decision.

That is the voice of nomologR.

Standard for methods

Every method in nomologR should be:

Research-backed — grounded in the methodological literature, with verifiable references in its documentation.
Situated historically — placed on the path from historical to contemporary practice. Historical techniques that learners will meet in published work (for example, eigenvalues greater than one, coefficient alpha, the Fornell–Larcker comparison, fixed fit-index cutoffs) may be shown as context, but they are labeled, qualified, and never silently substituted for contemporary evidence.
Honest about assumptions — the estimand, assumptions, and unsupported cases are stated; unsupported cases are refused or disclosed rather than approximated silently.
Demonstrable — shown first on data with a known answer, so learners can see what the evidence looks like when the truth is known.