TwinEdge AI Statistical Analysis

A comprehensive, browser-based statistical analysis suite built for engineers, analysts, and quality professionals. Powered by a high-performance WebAssembly engine, TwinEdge AI delivers professional-grade analytics directly in your browser — no installation required.

Quick Start

Get your first analysis result in under 60 seconds:

1. Load data. Click Sample Library and choose a dataset (e.g., "Student Health Survey"), or upload your own CSV/Excel file, or paste from the clipboard.
2. Pick an analysis. Open the Command Palette with Ctrl + K and type what you need — for example, "descriptive" or "t-test". Alternatively, use the category buttons in the header.
3. Configure & run. Select your columns and options in the dialog, then click Run. Results appear instantly in the right panel with formatted tables and charts.
4. Export. Click the export button on any result to download a PDF or Word report, or export the full session as a multi-page document.

Interface Tour

The interface is divided into four main areas:

Area	Location	Purpose
Header & Menu	Top	Category mega-menu, search, and navigation. Category buttons provide quick access to all 54+ analysis tools.
Worksheet	Left panel	Your data grid. Import, view, and edit data. Supports virtual rendering for large datasets. Click column headers to analyze.
Results Panel	Right panel	Session output with stacked analysis blocks. Each block shows formatted statistics and interactive charts. Modify & Re-Run any result.
Status Bar	Bottom	Row/column count, compute progress, and engine status (WebAssembly or TypeScript fallback).

The panel divider is draggable — resize the worksheet and results panels to suit your workflow.

Importing Data

Supported Formats

Format	Extensions	Notes
CSV	.csv, .tsv, .txt	Comma or tab delimited. Auto-detects separator.
Excel	.xlsx, .xls	Parsed via WebAssembly for speed. First sheet is loaded.
Clipboard	—	Paste tab-separated or comma-separated data directly.

Import Methods

• File Upload — Click the upload area or drag-and-drop a file.
• Clipboard Paste — Copy data from Excel or Google Sheets, then click the Paste button.
• Sample Library — Browse 13 curated datasets across 7 categories. Each sample includes metadata and suggested analyses.

Descriptive Statistics

Basic Stats

Compute a comprehensive set of summary statistics for one or more numeric columns, optionally grouped by a categorical variable.

Statistics Computed

Row	Measures
Location	N, N Missing, Mean, SE Mean, Trimmed Mean (5%), Mode
Spread	StDev, Variance, Range, IQR
Position	Minimum, Q1, Median, Q3, Maximum
Shape	Skewness, Kurtosis
Means	Geometric Mean, Harmonic Mean
Confidence Intervals	95% CI for Mean, Median, and StDev

How to Use

1. Open via menu Basic Stats → Descriptive Statistics or type "descriptive" in the Command Palette.
2. Select one or more numeric columns.
3. Optionally, choose a Group By variable to compute statistics for each group level.
4. Enable Custom Percentiles (checkbox) to add specific percentile calculations (e.g., 5, 10, 25, 50, 75, 90, 95).
5. Click Run. The output includes a professional formatted table plus histogram, boxplot, and dotplot.

Frequency Table

Basic Stats

Generate a frequency distribution table for a categorical or discrete variable, showing counts, percentages, and cumulative percentages.

How to Use

1. Open via Basic Stats → Frequency Table.
2. Select a column (text or numeric).
3. Click Run. The output shows each unique value with its count, percent, and cumulative percent.

Cross-Tabulation

Basic Stats

Create a two-way contingency table showing the joint frequency distribution of two categorical variables. Includes expected counts, row/column percentages, and a chi-square test of independence.

Correlation

Basic Stats

Compute correlation matrices to measure the strength and direction of linear relationships between numeric variables.

Methods Available

Method	Best For	Assumption
Pearson	Linear relationships	Normal distribution, continuous data
Spearman	Monotonic relationships	Ordinal or non-normal data
Kendall	Ordinal associations	Robust to ties and small samples
Partial	Controlling for confounders	Linear relationships after removing shared variance

Covariance Matrix

Basic Stats

Compute the variance-covariance matrix for two or more numeric variables.

Graphical Summary

Basic Stats

A comprehensive 4-panel graphical summary combining visualization and hypothesis testing in a single view — an industry-standard Graphical Summary layout.

Panels

Panel	Content
Top Left	Histogram with normal curve overlay
Top Right	Boxplot showing spread and outliers
Middle	95% confidence interval bars for Mean, Median, and StDev
Bottom	Anderson-Darling normality test results + descriptive statistics summary

Grubbs' Outlier Test

Basic Stats

Detect whether the most extreme value in a dataset is a statistically significant outlier.

Custom Percentiles

Basic Stats

Calculate specific percentile values for your data. Integrated into the Descriptive Statistics dialog — check the "Custom Percentiles" checkbox and enter comma-separated values (e.g., 5, 10, 25, 50, 75, 90, 95).

T-Tests

Hypothesis

Compare means using parametric t-tests. Three variants are available:

Z-Tests

Hypothesis

Compare means when the population standard deviation (σ) is known. More powerful than t-tests when σ is truly known.

Proportion Tests

Hypothesis

Test hypotheses about population proportions using the normal approximation (z-test).

Variance Tests

Hypothesis

Test hypotheses about population variances.

Normality Tests

Hypothesis

Assess whether your data follow a normal distribution. Four test methods are available:

Goodness-of-Fit Test

Hypothesis

Fit your data against multiple probability distributions and rank them by the Anderson-Darling statistic. Tests Normal, Exponential, Lognormal, and Uniform distributions simultaneously to find the best fit.

Output

• Distribution ranking — Sorted by AD statistic (lower is better).
• AD statistic and p-value for each candidate distribution.
• Parameter estimates (mean, σ, λ, etc.) for each fit.

Equal Variance Tests

Hypothesis

Test whether two or more groups have equal variances — a prerequisite for pooled t-tests and ANOVA.

Contingency Tests

Hypothesis

Test associations in categorical data using contingency tables.

Equivalence Testing (TOST)

Hypothesis

Use the Two One-Sided Tests (TOST) procedure to demonstrate that a parameter falls within a pre-specified equivalence margin. Essential for bioequivalence, quality assurance, and process validation.

Variants

• One-Sample TOST — Test if a sample mean is within [−Δ, +Δ] of a target.
• Two-Sample TOST — Test if the difference between two group means is within the equivalence bounds.
• Paired TOST — Equivalence test for paired observations.
• 2×2 Crossover Bioequivalence — Standard pharmaceutical crossover design with sequence, period, and treatment effects.

Power & Sample Size

Hypothesis

Determine the statistical power of a test or calculate the sample size needed to achieve a target power. Supports 16 test types plus tolerance intervals and CI-based sample size estimation:

Category	Test Types
Means	One-Sample T, Two-Sample T, Paired T, One-Sample Z, Two-Sample Z
Proportions	One Proportion, Two Proportions
Variances	One Variance, Two Variances
ANOVA	One-Way ANOVA
Equivalence	One-Sample TOST, Two-Sample TOST
Other	Poisson Rate, Tolerance Intervals, CI-Based Sample Size Estimation

The output includes a power curve chart showing how power changes with sample size or effect size.

Mann-Whitney U Test

Nonparametric

Compare two independent groups when normality assumptions are not met. Tests whether one distribution is stochastically greater than the other. Also known as the Wilcoxon rank-sum test.

Kruskal-Wallis Test

Nonparametric

The nonparametric alternative to one-way ANOVA. Tests whether three or more independent groups have the same distribution. Based on ranks rather than raw values.

Wilcoxon Signed-Rank Test

Nonparametric

The nonparametric alternative to the one-sample or paired t-test. Tests whether the median of a symmetric distribution equals a hypothesized value.

Sign Test

Nonparametric

A simple, distribution-free test for the population median. Counts the number of observations above and below the hypothesized median. Less powerful than Wilcoxon but makes fewer assumptions.

Mood's Median Test

Nonparametric

Test whether the medians of two or more groups are equal. More robust to outliers than Kruskal-Wallis.

Friedman Test

Nonparametric

The nonparametric alternative to two-way ANOVA (randomized block design). Tests whether the distributions of treatments are identical across blocks.

Runs Test

Nonparametric

Test whether a sequence of observations is random. Counts the number of runs (consecutive values above or below the median) and compares to the expected count under randomness.

Walsh Averages

Nonparametric

Compute pairwise Walsh averages (average of all pairs including self-pairs) for robust location estimation. The median of Walsh averages provides a robust, distribution-free estimator of the population center.

Theil-Sen Estimator

Nonparametric

A robust nonparametric regression estimator that computes the median of all pairwise slopes. Resistant to up to 29% outliers in the data.

OLS Regression

Modeling

Fit an ordinary least squares linear regression model with one or more predictors.

Output Includes

• Coefficient estimates with standard errors, t-values, and p-values
• Model summary: R², adjusted R², S (standard error of regression)
• ANOVA table for the overall model (F-test)
• Residual diagnostics: four-pack plot (residuals vs fits, normal probability plot, histogram, residuals vs order)
• Fitted line plot (for single predictor models)

Stepwise Regression

Modeling

Automatically select the best set of predictors using forward selection, backward elimination, or bidirectional stepwise procedures. Controlled by α-to-enter and α-to-remove thresholds.

Best Subsets Regression

Modeling

Evaluate all possible combinations of predictors and rank models by R², adjusted R², Mallows' Cp, and BIC.

Polynomial Regression

Modeling

Fit polynomial models (quadratic, cubic, etc.) to capture nonlinear relationships. Specify the polynomial degree and get coefficients, R², lack-of-fit test, and a fitted curve plot.

Logistic Regression

Modeling

Fit binary, ordinal, or nominal logistic regression models for categorical outcomes.

Poisson Regression

Modeling

Model count data using the Poisson distribution with a log link function. Reports rate ratios, deviance, Pearson chi-square, and overdispersion test.

Nonlinear Regression

Modeling

Fit nonlinear models using the Levenberg-Marquardt algorithm. Six built-in model forms:

• Exponential Growth/Decay — y = a · e^bx
• Power — y = a · x^b
• Logistic (4P) — Four-parameter logistic curve
• Gaussian — Bell curve fit
• Michaelis-Menten — Enzyme kinetics model
• Asymptotic — y = a · (1 − e^−bx)

Partial Least Squares (PLS)

Modeling

PLS regression using the NIPALS algorithm. Ideal for datasets with many correlated predictors or when the number of predictors exceeds the number of observations. Reports component loadings, R² per component, and VIP scores.

Orthogonal / Deming Regression

Modeling

Regression for method comparison when both X and Y have measurement error. Minimizes perpendicular distances rather than vertical residuals. Specify the variance ratio (λ) or use Deming regression (λ = 1).

Stability Studies

Modeling

ICH-guideline stability analysis for pharmaceutical shelf-life estimation. Fits linear regression to stability data (time vs. response), tests against specification limits, and estimates the time at which the 95% confidence bound crosses the spec limit.

One-Way ANOVA

ANOVA

Test whether the means of three or more groups are equal. Partitions total variation into between-group and within-group components.

Output Includes

• ANOVA table with SS, DF, MS, F-statistic, and p-value
• Group means, standard deviations, and 95% CIs
• Individual value plot and boxplots by group
• Interval plot with confidence intervals
• Optional post-hoc tests (Tukey, Fisher LSD, Dunnett, Bonferroni, Games-Howell, Sidak)

Two-Way ANOVA

ANOVA

Analyze the effects of two factors and their interaction on a continuous response. Produces an ANOVA table with main effects and interaction terms, plus main effects and interaction plots.

General Linear Model (GLM)

ANOVA

The most flexible ANOVA option. Handles unbalanced designs, multiple factors, covariates (ANCOVA), and nested terms. Uses Type III sums of squares by default.

Welch ANOVA

ANOVA

A robust alternative to standard one-way ANOVA when group variances are unequal. Uses Welch's F-test and Games-Howell post-hoc comparisons that do not assume equal variances.

Nested ANOVA

ANOVA

Analyze hierarchical/nested designs where levels of one factor are nested within levels of another (e.g., machines nested within factories). Decomposes variance into between-group, within-group (nested), and residual components.

Balanced ANOVA

ANOVA

Analysis of variance for balanced (equal sample sizes) multi-factor designs. Provides exact Type I sums of squares and clean variance decomposition.

ANCOVA (Analysis of Covariance)

ANOVA

Combine ANOVA with regression by including one or more continuous covariates. Tests for group differences after adjusting for the covariate, improving precision and controlling for confounding variables.

Repeated Measures ANOVA

ANOVA

Analyze data where the same subjects are measured under multiple conditions or across time. Accounts for within-subject correlation using the Greenhouse-Geisser or Huynh-Feldt epsilon correction when the sphericity assumption is violated.

Mixed Effects ANOVA

ANOVA

Models with both fixed and random effects. Appropriate for designs where some factors represent fixed treatment levels and others are random samples from a population (e.g., operators, batches, machines). Estimates variance components for random effects.

Analysis of Means (ANOM)

ANOVA

Compare each group mean against the overall mean using decision limits. Unlike ANOVA (which tests whether any means differ), ANOM identifies which specific groups are significantly above or below the grand mean. Produces an ANOM chart with decision lines.

Post-Hoc Comparisons

ANOVA

After a significant ANOVA result, determine which specific groups differ:

Control Charts

Quality

Monitor process stability using Statistical Process Control (SPC) charts. TwinEdge AI includes 19 control chart types across six categories:

Variables Charts

Chart	Use Case
I-MR	Individual observations + moving range. Most common for continuous data.
Xbar-R	Subgroup means + ranges. For subgroup sizes 2–10.
Xbar-S	Subgroup means + standard deviations. For subgroup sizes >10.
I-MR-R/S	Between/Within chart for subgroup data. Three panels: I, MR, and R or S.
Zone Chart	Zone-based scoring chart. Assigns weighted scores by zone (A/B/C).
Moving Average	Smoothed moving average with moving range chart.

Attributes Charts

Chart	Use Case
P Chart	Proportion defective. Variable sample sizes allowed.
NP Chart	Number defective per fixed sample size.
C Chart	Defects per unit (equal inspection units).
U Chart	Defects per unit with variable sample sizes.
Laney P'	Adjusted P chart for overdispersed proportion data.
Laney U'	Adjusted U chart for overdispersed rate data.

Time-Weighted Charts

Chart	Use Case
EWMA	Exponentially weighted moving average. Sensitive to small sustained shifts.
CUSUM	Cumulative sum. Detects small persistent process shifts.

Rare Events & Multivariate

Chart	Use Case
G Chart	Geometric chart for count of events between rare occurrences.
T Chart	Time between rare events chart.
Hotelling's T²	Multivariate control chart for monitoring multiple correlated quality characteristics simultaneously.
Run Chart	Simple time-order plot with median line and runs tests for randomness.

Each chart displays the center line (CL), upper control limit (UCL), and lower control limit (LCL) with out-of-control points highlighted.

Capability Analysis

Quality

Assess how well a process meets specification limits. Six capability study types are available:

Measurement System Analysis

Quality

Evaluate and qualify measurement systems using 7 MSA study types:

Scatterplot

Graphs

Create an X-Y scatterplot to visualize the relationship between two numeric variables.

Time Series Plot

Graphs

Plot numeric values in observation order to identify trends, cycles, and patterns over time.

Pareto Chart

Graphs

A bar chart sorted by frequency (descending) with a cumulative percentage line. Follows the 80/20 rule.

New Chart Types

Graphs

TwinEdge AI now includes a full suite of standalone graph types you can create from the Graphs menu:

All Chart Types

Graphs

TwinEdge AI includes 30+ chart types that are generated as part of analysis results or created standalone:

Autocorrelation (ACF & PACF)

Time Series

Compute and plot the sample autocorrelation function (ACF) and partial autocorrelation function (PACF) with 95% confidence bounds. Essential for identifying the order of ARIMA models.

Cross-Correlation

Time Series

Compute the cross-correlation function (CCF) between two time series at positive and negative lags. Identifies leading/lagging relationships between variables.

Smoothing & Trend Analysis

Time Series

Decomposition

Time Series

Decompose a time series into trend, seasonal, and residual components using additive or multiplicative models. Output includes a 4-panel chart (original, trend, seasonal, residual) and component data.

ARIMA

Time Series

Fit ARIMA(p,d,q) models for stationary and non-stationary time series. Specify the autoregressive (p), differencing (d), and moving average (q) orders. Output includes parameter estimates, residual diagnostics, fitted vs actual plot, and forecasts with confidence intervals.

Factorial Designs

DOE

Create and analyze factorial experimental designs to study the effects of multiple factors on a response.

Factorial Analysis Output

• Effects table with coefficients, standard errors, T-values, and p-values
• ANOVA table (SS, DF, MS, F, p)
• Pareto chart of standardized effects with significance line
• Normal probability plot of effects (half-normal)
• Main effects and interaction plots
• R², adjusted R², and predicted R²

Screening Designs

DOE

Response Surface Methodology

DOE

Optimize process parameters by fitting second-order (quadratic) models to experimental data.

Principal Component Analysis (PCA)

Multivariate

Reduce dimensionality by extracting principal components that explain the maximum variance. Output includes eigenvalues, proportion of variance explained, scree plot, loading plot, biplot, and score matrix.

Factor Analysis

Multivariate

Identify latent factors underlying observed variables. Supports Varimax and Quartimax rotation methods. Reports factor loadings, communalities, uniqueness, and rotated factor matrix.

Clustering

Multivariate

Discriminant Analysis (LDA)

Multivariate

Linear Discriminant Analysis for classifying observations into known groups. Reports discriminant functions, classification accuracy, confusion matrix, and group centroids.

Correspondence Analysis

Multivariate

Visualize associations in contingency tables by mapping row and column categories to a low-dimensional space. Produces a symmetric biplot showing the relationships between categories.

MANOVA

Multivariate

Multivariate Analysis of Variance for testing group differences across multiple response variables simultaneously. Reports Wilks' Lambda, Pillai's Trace, Hotelling-Lawley Trace, and Roy's Largest Root test statistics.

Item Analysis

Multivariate

Assess the reliability of multi-item scales using Cronbach's Alpha. Reports overall alpha, item-total correlations, and alpha-if-item-deleted for each item.

Classification & Regression Trees (CART)

Predictive

Build decision trees for classification or regression. Output includes a visual tree diagram, variable importance rankings, confusion matrix (classification), R² (regression), and cross-validation accuracy.

Ensemble Methods

Predictive

K-Nearest Neighbors (KNN)

Predictive

Classify or predict by finding the K nearest observations in feature space. Supports classification (majority vote) and regression (mean of neighbors). Reports accuracy, confusion matrix, and optimal K selection.

Support Vector Machine (SVM)

Predictive

Maximum-margin classifier with linear or RBF (radial basis function) kernel. Reports support vector count, classification accuracy, ROC curve (binary), and decision boundary visualization.

Neural Network (MLP)

Predictive

Multi-layer perceptron with configurable hidden layers and backpropagation training. Reports training loss history, classification/regression metrics, confusion matrix, and ROC curve.

Rank & Standardize

Data

Rank

Convert numeric values to their rank order (1, 2, 3, ...). Tied values receive the average rank.

Standardize

Compute z-scores: (x − mean) / std_dev. Centers data at 0 with unit standard deviation.

Worksheet Operations

Data

A suite of worksheet-level operations for reshaping and combining data:

Column Transforms

Data

Apply transformations to create or modify columns:

• Recode — Map old values to new values (e.g., 1→"Low", 2→"Medium", 3→"High").
• Indicator Variables — Create 0/1 dummy columns from a categorical variable.
• Change Data Type — Convert between numeric and text representations.
• Concatenate — Join two or more text columns into a single column with a separator.

Generate Data

Data

Create new columns with patterned or random data. Distributions available: Normal, Uniform, Poisson, Binomial, Exponential, and patterned sequences. Useful for simulation, testing, and demonstration.

Column / Row Statistics

Data

Compute column-wise or row-wise summary statistics and store the result as a new column. Supports sum, mean, median, min, max, count, standard deviation, and more.

Find & Replace

Data

Search and replace values within a column. Supports exact match, contains, and regular expression patterns. Preview matches before applying changes.

Date/Time Extraction

Data

Extract components from date/time text columns: year, month, day, hour, minute, second, day of week, and week number. Creates new numeric columns for each extracted component.

Formula Engine

Data

Create computed columns using the formula bar. Enter an expression referencing existing column names to generate new calculated columns. Runs via WebAssembly for fast evaluation.

Edit Mode & Graph Editing

Before exporting, use Edit Mode to customize your results for publication-ready reports. Available on every result block:

Feature	Description
Editable Title	Click any chart block title to rename it inline. Custom titles are saved and appear in exports.
Add Notes	Per-block sticky-note editor — notes are included verbatim in PDF and Word exports with styled callout formatting.
Exclude from Export	Toggle a block to be included or excluded from full-report exports.
Hide/Show Sections	Each MonoBlock section (tables, statistics) has a Visible/Hidden toggle — hidden sections are absent from export.
Reference Lines	Add horizontal or vertical reference lines to any chart with a color picker (6 colors), optional label text, and a manage/remove panel.
Block Reordering	Use arrow-up/arrow-down buttons to reorder blocks in the report layout.
Modify & Re-Run	Re-open any analysis dialog pre-filled with previous parameters to adjust and re-run instantly.

PDF Reports

Export analysis results as professionally formatted PDF documents.

Export Options

• Single Result — Export an individual analysis block to PDF.
• Full Report — Export the entire session as a multi-page report with a title page and table of contents.

PDF reports include monospace-formatted statistical output, high-resolution chart images, branded headers/footers, page numbers, and timestamps.

Word (DOCX) Reports

Export analysis results as Microsoft Word documents with the same clean styling as PDF reports.

Export Options

• Single Result — Export an individual analysis block to Word.
• Full Report — Export the entire session as a multi-section Word document with a title page, table of contents, branded headers/footers, and page numbers.

Word reports use the same layout as PDFs: emerald-branded headers, alternating-row monospace output, bordered chart images, and professional footers. Edit, annotate, or incorporate results into existing Word documents.

Store to Worksheet

After running Descriptive Statistics, click the "Store to Worksheet" button to save computed statistics back to your data as new columns.

Copy Results

Every result block has a Copy button that copies the monospace-formatted output to your clipboard. Paste directly into reports, emails, or documentation.

Keyboard Shortcuts

Shortcut	Action
Ctrl + K	Open Command Palette — search and launch any analysis
↑ ↓	Navigate items in Command Palette and dialogs
Enter	Select highlighted item or confirm action
Esc	Close current dialog, palette, or overlay
Tab	Move between form controls in dialogs
Ctrl + Z	Undo last analysis block
Ctrl + Shift + C	Copy active result to clipboard
Alt + 1–9	Switch to block tab N (block navigation)

Sample Datasets

The Sample Library includes 41 curated datasets across 12 categories to help you explore every feature:

Basic Statistics (3 datasets)

Dataset	Size	Best For
Student Health Survey	200 × 12	Descriptive stats, t-tests, correlation
Edge Case Validator	10 × 6	Stress-test values (constants, NaN, negatives)
Categorical Survey Data	60 × 6	Frequency, cross-tab, demographics + Likert

Hypothesis Tests (6 datasets)

Dataset	Size	Best For
Before / After Treatment	12 × 3	Paired t-test, sign test, Wilcoxon
Drug Trial Proportions	40 × 2	Proportion z-tests
Machine Precision Comparison	30 × 2	Variance tests (F-test, Bonett)
Distribution Shape Sampler	100 × 4	Normality tests, goodness-of-fit
Bioequivalence Crossover	24 × 5	TOST, 2×2 crossover bioequivalence
Sample Size Planning	20 × 3	Power analysis demonstration

Regression / Modeling (5 datasets)

Dataset	Size	Best For
Multi-Variable Regression	20 × 5	OLS, stepwise, best subsets
Polynomial Growth Curve	30 × 2	Polynomial, nonlinear regression
Manufacturing Defect Counts	60 × 4	Poisson regression
Drug Stability Study (ICH)	36 × 3	ICH stability, shelf-life estimation
Clinical Outcome Prediction	80 × 6	Binary logistic regression

ANOVA (4 datasets)

Dataset	Size	Best For
ANOVA Test Groups	19 × 4	One-way/two-way ANOVA, post-hoc
Nested Factory Study	36 × 3	Nested ANOVA, variance components
Drug Efficacy with Covariate	30 × 3	ANCOVA, covariate adjustment
Unequal Variance Groups	45 × 2	Welch ANOVA, ANOM

Quality Tools (7 datasets)

Dataset	Size	Best For
SPC Monitoring Data	26 × 5	Control charts, I-MR, Xbar-R
Gage R&R Study	30 × 4	Measurement system analysis
Subgroup Control Data	100 × 5	Xbar-S, subgroup control charts
Precision Diameter Process	100 × 1	Capability analysis (Cp, Cpk)
Attribute Agreement Study	90 × 5	Attribute agreement, Kappa
Rare Event Tracking	30 × 2	G chart, T chart
Multivariate Process Control	50 × 3	Hotelling T² chart

Nonparametric (2 datasets)

Dataset	Size	Best For
Rating Scale Data	15 × 4	Friedman, sign test, Wilcoxon
Multi-Group Nonparametric	15 × 3	Mann-Whitney, Kruskal-Wallis, Mood's Median

Graphs (5 datasets)

Dataset	Size	Best For
Biometric Scatter Data	40 × 4	Scatter, bubble, marginal plots
Defect Category Data	20 × 2	Bar, pie, Pareto charts
Monthly Revenue Series	36 × 2	Area, ECDF, symmetry plot
3D Surface & Contour Data	40 × 3	3D surface, contour plots
Multi-Vari Process Data	36 × 4	Multi-Vari chart

Time Series (2 datasets)

Dataset	Size	Best For
Airline Passengers (Monthly)	48 × 2	Decomposition, ARIMA, smoothing
Daily Sales & Temperature	90 × 3	Cross-correlation, trend analysis

DOE (2 datasets)

Dataset	Size	Best For
2&sup4; Factorial Design	32 × 5	Factorial analysis, effects plots
CCD Response Surface	23 × 4	RSM analysis, optimization

Multivariate (2 datasets)

Dataset	Size	Best For
Iris-Style Measurements	45 × 7	PCA, clustering, LDA
Product Quality Survey	35 × 10	Factor analysis, item analysis

Predictive Analytics (1 dataset)

Dataset	Size	Best For
Customer Churn Dataset	50 × 11	CART, random forest, SVM, neural net

Large Datasets (3 datasets)

Dataset	Size	Best For
Manufacturing 100K	100K × 15	Performance testing
Manufacturing 500K	500K × 15	Stress testing
Manufacturing 1M	1M × 15	Maximum-scale testing

Spreadsheet Themes

Customize the worksheet appearance with 6 built-in themes. Click the palette icon in the worksheet header to switch.

Theme	Style
Default	Light gray with violet accent
Ocean	Sky blue headers
Forest	Mint green tones
Slate	Cool gray, minimal
Rose	Soft pink accents
Amber	Warm yellow highlights

Session Management

TwinEdge AI automatically saves your work and provides tools for managing sessions:

Feature	Description
Auto-Save	Session state (data, analysis blocks, annotations, panel width) is automatically saved to browser storage every 500ms. Reload the page to restore.
New Session	Clear all data and analysis blocks to start fresh.
Export Session	Download the entire session state as a JSON file for archiving or sharing.
Import Session	Upload a previously exported JSON session file. The page reloads with the restored state.

Performance Tips

• Use WebAssembly — Ensure the status bar shows "WASM" for maximum performance.
• Column-click analysis — Clicking a column header runs descriptive stats directly without opening a dialog.
• Subset first — For very large datasets, use Subset Rows to filter to relevant data before running analyses.
• Close unused results — Remove completed analysis blocks you no longer need.
• Session auto-save — Your work is automatically saved to browser storage. Reload to restore.

Full Tool Index

All 86+ analysis tools at a glance, organized by category:

Basic Statistics (7 tools)

Tool	Description
Descriptive Statistics	Summary stats, histogram, boxplot, CIs, percentiles
Frequency Table	Counts and percentages for categorical data
Cross-Tabulation	Two-way contingency table with chi-square
Correlation	Pearson, Spearman, Kendall, Partial matrices
Covariance	Variance-covariance matrix
Graphical Summary	4-panel visual + normality test
Grubbs' Test	Single outlier detection

Hypothesis Tests (11 tools)

Tool	Description
T-Tests (1-sample, 2-sample, paired)	Parametric mean comparisons
Z-Tests (1-sample, 2-sample)	Mean tests with known σ
Proportion Tests (1-prop, 2-prop)	Binomial proportion z-tests
1-Variance / 2-Variances	Chi-square and F-test / Bonett for variances
Equal Variance	Levene's and Bartlett's tests
Normality Tests	Anderson-Darling, Shapiro-Wilk, Ryan-Joiner, KS
Goodness-of-Fit	Multi-distribution fit ranking by AD statistic
Contingency Tests	Chi-Square and Fisher's Exact
Equivalence (TOST)	1-sample, 2-sample, paired, 2×2 crossover
Power & Sample Size	16 test types + tolerance intervals + CI estimation
Nonparametric Tests	Mann-Whitney, Kruskal-Wallis, Wilcoxon, Sign, Mood, Friedman, Runs, Walsh, Theil-Sen

Regression & Modeling (11 tools)

Tool	Description
OLS Regression	Linear regression with diagnostics
Stepwise	Forward, backward, and bidirectional variable selection
Best Subsets	All-subsets model comparison (R², Cp, BIC)
Polynomial	Quadratic, cubic, and higher-order fits
Binary Logistic	Binary outcome regression with odds ratios
Ordinal Logistic	Ordered categorical outcomes
Nominal Logistic	Unordered categorical outcomes
Poisson	Count data regression (log link)
Nonlinear	6 model forms via Levenberg-Marquardt
PLS (Partial Least Squares)	NIPALS algorithm for correlated predictors
Orthogonal / Deming	Method comparison with error in both variables
Stability Studies	ICH-guideline shelf-life estimation

ANOVA (10 types + 7 post-hoc methods)

Tool	Description
One-Way ANOVA	Single-factor analysis of variance
Two-Way ANOVA	Two-factor with interaction
General Linear Model	Unbalanced designs, covariates, nested terms
Welch ANOVA	Robust to unequal variances
Nested ANOVA	Hierarchical/nested designs
Balanced ANOVA	Equal sample size multi-factor designs
ANCOVA	ANOVA with continuous covariates
Repeated Measures	Within-subject designs with sphericity correction
Mixed Effects	Fixed + random effects with variance components
ANOM	Analysis of means with decision limits
Post-Hoc (7 methods)	Tukey, Fisher LSD, Dunnett, Bonferroni, Games-Howell, Sidak, Hsu MCB

Quality Tools (32 tools)

Tool	Description
Control Charts (19+ types)	I-MR, Xbar-R/S, I-MR-R/S, Zone, MA, P, NP, C, U, Laney P'/U', EWMA, CUSUM, G, T, Hotelling T², Run
Capability (6 types)	Normal, Sixpack, Non-Normal, Between/Within, Multi-Variable, Perf vs Spec
MSA (7 types)	Crossed R&R, Nested R&R, Linearity & Bias, Expanded R&R, Attribute Agreement, Attribute Analytic, Type 1

Graphs (15 standalone types)

Tool	Description
Scatterplot	X-Y scatter with optional regression line
Time Series Plot	Values in observation order
Pareto Chart	Sorted frequency bars with cumulative line
Bar Chart	Categorical variable frequencies
Pie Chart	Proportional distribution with labels
Area Chart	Filled area trend visualization
Bubble Plot	3-variable scatter with size encoding
Stem-and-Leaf	Text-based distribution display
Marginal Plot	Scatter with marginal histograms
Empirical CDF	Cumulative distribution step function
Symmetry Plot	Median symmetry assessment
Contour Plot	Filled isolines for 3-variable data
Multi-Vari Chart	Variation across categorical factors
3D Surface Plot	Interactive wireframe with rotation
3D Scatterplot	Interactive 3D point cloud

Time Series & Forecasting (8 tools)

Tool	Description
Autocorrelation (ACF)	Sample ACF with confidence bounds
Partial ACF (PACF)	Durbin-Levinson recursion
Cross-Correlation	CCF between two series at ± lags
Moving Average	Centered/trailing MA with accuracy measures
Trend Analysis	Linear, quadratic, growth, S-curve with forecasts
Decomposition	Additive/multiplicative trend + seasonal + residual
Exponential Smoothing	Single, Holt's, Winters' with forecasts
ARIMA	ARIMA(p,d,q) fitting with diagnostics and forecasts

DOE — Design of Experiments (7 tools)

Tool	Description
2k Full Factorial	Complete design with all factor combinations
2k-p Fractional Factorial	Screening designs with confounding
Plackett-Burman	Economical screening for main effects
General Full Factorial	Multi-level mixed factorial designs
Central Composite (CCD)	Response surface design for optimization
Box-Behnken	Response surface without corner points
RSM Analysis	Quadratic model fitting with surface/contour plots

Multivariate Analysis (8 tools)

Tool	Description
Principal Components (PCA)	Dimensionality reduction with scree/biplot
Factor Analysis	Latent factors with Varimax/Quartimax rotation
K-Means Clustering	Partition into K clusters with silhouette scores
Hierarchical Clustering	Agglomerative with dendrogram
Discriminant Analysis	Linear classification with confusion matrix
Correspondence Analysis	Contingency table biplot visualization
MANOVA	Multivariate ANOVA (Wilks, Pillai, Hotelling, Roy)
Item Analysis	Cronbach's Alpha reliability

Predictive Analytics / ML (6 tools)

Tool	Description
CART (Decision Tree)	Classification/regression tree with visual diagram
Random Forest	Ensemble with bagging and feature subsampling
Gradient Boosting	Sequential boosted trees with loss curves
K-Nearest Neighbors	Instance-based classification/regression
Support Vector Machine	Max-margin classifier (linear/RBF kernel)
Neural Network (MLP)	Multi-layer perceptron with backpropagation

Data Operations (7 tools)

Tool	Description
Rank / Standardize	Rank order or z-score transformation
Worksheet Operations	Stack, unstack, subset, transpose, sort, merge
Column Transforms	Recode, indicator variables, change type, concatenate
Generate Data	Random/patterned data from distributions
Column/Row Statistics	Row-wise or column-wise summary as new column
Find & Replace	Search/replace with regex support
Date/Time Extraction	Extract year, month, day, hour from date text

---

TwinEdge AI Statistical Analysis — Help Guide
© 2026 TwinEdge AI. All rights reserved.