SEMinR 2.5.0 Released
SEMinR v2.5.0 is now on CRAN, bringing prediction to every interaction type, more flexible plots, a public accessor API for developers, and a round of bug fixes.
SEMinR 2.5.0 is released on CRAN! This release brings prediction to every interaction type, more flexible plots, and a round of bug fixes.
Improvements to prediction: predict() and predict_pls() now work with ALL interaction methods — product_indicator, orthogonal, and quadratic terms — not just two-stage. Single predictions, k-fold cross-validation, and LOOCV all work across the board, and k-fold CV can now run in parallel for a real speed-up on larger jobs.
New plotting feature: Custom confidence levels in plots — show 90%, 99%, or any alpha instead of the fixed 95%.
Developer-friendly tools: A public, documented accessor API for constructs and measurement-model elements, so you can stop reaching into seminr::: for internals.
Lots of fixes:
- Bootstrapping now works with PLSc + higher-order constructs
- Fixed the parallel-bootstrap “no package called seminr” error
interaction_term()no longer errors on a single-IV quadratic- CBSEM summaries now report path directions correctly (IV → DV)
- Plot stars and summary-table p-values now agree
A big thank you to the contributors behind this release:
- Nicholas Danks — prediction support for all interaction methods, plus the plot/summary p-value consistency fix
- Andrej Pawluczenko — the custom confidence-level plotting feature
Update now:
install.packages("seminr")