metis Updates | PLS-SEM App Release Notes

v0.3.1 latest release August 2026

metis 0.3.1 introduces a centralized Preferences and Algorithm Defaults hub, an interactive missing-value finder in DataView, missing-value marker with defaults value and a custom one you can add, full support for Higher-Order Construct estimation and Bootstrapping, unified PLSpredict and CVPAT prediction summaries, and complete multilingual translation across English, Spanish, Portuguese, and French.

Preferences & Algorithm Defaults

Centralized Settings Hub: Set your favorite analysis defaults once in a clean, dedicated preferences menu without hunting through menus before every run.
Metis & Game: Get engaging with metis 2048 game during calculation time; turn it on in settings. This does not affect the backend calculation but creates an engaging environment between the rsearcher and the research.

DataView Missing-Value Navigator

Interactive Missing-Value Counter: The data toolbar now features a convenient button displaying exactly how many empty or missing cells were detected upon import.
1-Click Cycle Navigation: Click “Find next” or “Find previous” to smoothly glide across thousands of rows and columns directly to each missing entry.
Illuminated Crosshair Highlighting: The active cell, row header, and column header illuminate with a crisp accent highlight so you can instantly spot where data needs attention.
Safe & Non-Destructive: Quickly discover and verify data gaps without accidentally overwriting values or altering your dataset layout.
Missing-Value Marker Define your missing value marker using Metis default missing-value markers or create a custom missing-value marker and save for future use.

Higher-Order Construct (HOC) Modeling

Multiple Estimation Methods: Easily estimate multidimensional concepts using either Repeated Indicators or Two-Stage (Embedded and Disjoint) approaches.
Seamless Canvas & Bootstrap Integration: Choose your higher-order approach directly when configuring your model, with full support during Bootstrap resampling.

PLSpredict & CVPAT Evaluation

Authoritative Prediction Summaries: Clean, easy-to-read tables bring together Q²predict, PLS-SEM prediction errors (RMSE and MAE), and benchmark comparisons in one place.
Dual-Section CVPAT Analysis: Directly compare your PLS-SEM model’s predictive power against both standard Linear Models and simple Indicator Averages at indicator and construct levels.

Complete Multilingual Experience

Four Fully Supported Languages: Switch the entire application interface, settings, tooltips, and analysis options seamlessly between English, Spanish (Español), Portuguese (Português), and French (Français).
Thoughtfully Adapted Layouts: Dynamic label spacing ensures all statistical terms and descriptions remain natural, elegant, and readable in every supported language.

v0.0.1beta May 2026

metis 0.0.1beta brings the full PLS-SEM workflow into one desktop app: setup, workspaces, dataset handling, visual model building, PLS-SEM, Bootstrap, PLSpredict, Advanced analysis, results review, exports, and TARK reporting.

Workspaces & Data

Workspace Home now manages workspaces, models, datasets, and saved results from one project browser.
Single-file .ada workspaces embed datasets, support bundled samples, and open directly from File Explorer through the metis file association.
Dataset import supports spreadsheet and text data, then opens into DataView for preview, editing, replace, or save-as-new workflows.
Descriptive statistics and dataset persistence keep imported data available across model, results, and reporting screens.

Model Canvas

Build reflective and formative measurement models with draggable constructs, indicators, measurement paths, and structural paths.
Create moderation relationships by connecting a construct to an existing path, with dotted moderation links anchored to the target relationship.
Work across multi-model tabs with open, close, reorder, same-session switching, per-model dirty state, and active dataset syncing.
Use multi-select drag, shared resize handles, alignment guides, snap controls, copy, cut, paste, select all, Save As, Fit to Screen, and PNG export.
Autosave and Run on Draw reduce lost work and make quick model checks available while the canvas is being built.

Analysis Engine

Run PLS-SEM through the R backend with seminr, including Standard PLS, Consistent PLS, reflective constructs, and formative constructs.
Run Bootstrap analyses with path, indirect, total effect, loading, weight, confidence interval, and significance outputs.
Run PLSpredict with configurable folds, repetitions, optional CVPAT, Q²predict, PLS-SEM RMSE/MAE, and linear-model comparison metrics.
Run Advanced analysis after a valid PLS-SEM result, including IPMA, NCA, and cIPMA target-construct diagnostics powered by seminrExtras.

Results & Diagnostics

Results View has mode-specific sidebars for PLS-SEM, Bootstrap, PLSpredict, and Advanced analysis.
Core panels cover path coefficients, total effects, specific indirect effects, outer loadings, outer weights, reliability, discriminant validity, cross-loadings, R², f², VIF, model fit, and execution logs.
Bootstrap panels separate resampled structural and measurement effects from base-model reference panels, including visible reference-state labels.
PLSpredict panels include MV/LV summaries, prediction errors, PLS vs LM comparison, Q²predict, error histograms, and CVPAT states.
Advanced analysis panels include priority maps, construct tables, necessity checks, ceiling lines, bottleneck tables, and cIPMA priorities.
The results diagram supports metric overlays, hoverable construct diagnostics, result-side movement, and cleaner export behavior.

Reporting & Exports

Saved result artifacts reopen PLS-SEM, Bootstrap, PLSpredict, and Advanced analysis outputs from the workspace.
Result tables can be copied as rich text, exported to Excel, or included in a clean HTML report with copy-table controls.
R scripts can be copied or exported so the analysis can be inspected and reproduced outside the desktop app.
TARK turns saved PLS-SEM, Bootstrap, and PLSpredict results into journal-ready report tables, with optional Advanced analysis families.
TARK report families cover measurement model assessment, discriminant validity, structural model assessment, explanatory and predictive power, model fit, PLSpredict, IPMA, NCA, and additional effects.

Setup, Packaging & Security

Lite and Bundle beta installers are supported: Lite connects to an existing R installation, while Bundle ships the portable R runtime.
The setup wizard validates R, installs required packages, and checks seminr, seminrExtras, semPower, and supporting runtime packages.
Installer and setup windows now use stable packaged renderer loading, opaque shell styling, and corrected file-protocol privileges.
Security hardening restricts Electron IPC, approved file paths, export opening, local Plumber access, dataset roots, and R executable selection.
Production builds remove generated debug files and avoid leaking local development paths into shipped artifacts.
Light/dark themes, preferences, title-bar actions, onboarding tour, diagnostics, and clearer error messaging round out the beta experience.

Beta Notes

Known beta limitations remain documented in the user docs, and future public updates will build from this single beta baseline.