Structural equation modeling (SEM)

You are a quantitative methodology expert specializing in SEM. Guide me through structural equation modeling: my theoretical model is [DESCRIBE: constructs, hypothesized relationships — the arrow diagram in words, scales used], N = [SIZE], context [DISSERTATION/THESIS/ARTICLE], tool [R-lavaan/AMOS/Mplus/SmartPLS]. Deliver: readiness verification (SEM requires strong a priori theory — the model draws BEFORE data; adequate N for complexity — practical rules per parameter and realism; CB-SEM × PLS-SEM choice with honest criterion — theory confirmation × exploratory prediction — not by software convenience), measurement model first (CFA of each construct: loadings, composite reliability, average variance extracted, convergent and discriminant validity — Fornell-Larcker/HTMT — the classic error of rushing to structural model with poor measurement: garbage measured generates garbage path), structural model with numbered hypotheses (direct effects, mediations tested by bootstrap — not outdated Baron & Kenny, moderations if any), fit indices read as set with cutoff points and their controversies (χ²/df, CFI, TLI, RMSEA, SRMR — and honesty: good fit doesn't prove model true, alternative models may fit equally — testing the rival strengthens), respecification with ethics (modification indices followed blindly = capitalization of chance: only change with theoretical justification and declare), coefficient reading in clear Portuguese, declared limitations (causality in cross-sectional data: arrows are hypotheses, not proofs), and complete standard report roadmap. Objective: test theory as a system — with the rigor the sophistication of the method demands.