Statistical Lab
A statistical analysis platform that takes you from raw data to a publication-ready result — in six steps, without writing code.
Overview
Statistical Lab is organized into two modes. Explore is for understanding data and testing hypotheses — it holds Exploration, Comparison, Relationship, and Econometrics. Model is for prediction and structure — it holds Predictive, Clustering, Time Series, and Structural. Switch modes from the top of the sidebar to see the relevant analyses.
Not sure which method to use? Open Analysis Recommendation from the dashboard — it reads your dataset and suggests the most suitable Statistical Lab analysis, then sends you straight to it.
Explore & Model modes
68+ Statistical Methods
From descriptive statistics to SEM, clustering, and time series — all in one place.
AI-powered results
Every analysis produces a plain-language Summary, a Reasoning tab, and a full APA report.
Reproducible code
Export Python or R code for any analysis — runnable in your own environment.
What you need
What you get
Analysis categories
Use the left sidebar to navigate between categories. Upload your data at the top, search for a specific method, or browse by category. Each analysis page includes a format guide and a sample dataset to get started immediately.
Sidebar navigation & analysis guide
Statistical Lab organizes its 68+ methods into 8 categories by research goal.
Exploration
ExploreUnderstand your data before testing.
3 methods
Comparison
ExploreTest whether groups differ.
13 methods
Relationship
ExploreModel how variables predict each other.
10 methods
Econometrics
ExploreEstimate causal effects from observational data.
4 methods
Predictive
ModelClassify or forecast new observations.
12 methods
Structural
ModelUncover latent constructs and path models.
9 methods
Clustering
ModelFind natural groups in your data.
6 methods
Time Series
ModelAnalyze and forecast time-ordered data.
11 methods
Running an analysis
Each analysis page opens with a summary card — what it measures, when to use it, requirements, and what you'll learn. Load the sample dataset to see the full workflow before using your own data.
An analysis page with a sample dataset
Uploading your own data
When you're ready to use your own dataset, just drag a file anywhere onto the page. A full-screen drop zone appears in the center — release the file there and it uploads instantly. CSV and Excel (.xlsx / .xls) are supported. You can also use the upload control at the top of the sidebar.
데이터를 먼저 정리해야 한다면 데이터 준비 가이드를 확인하세요. 결측치 처리, 이상치 제거, 변수 변환 등 전처리를 마친 뒤 Send to Statistical Lab 버튼으로 바로 연결됩니다.
Drag & drop to upload
Every method follows the same 6-step workflow. Steps 1–3 cover selecting variables, configuring settings, and validating your data.
Select the columns your analysis will use.
Configure parameters — sensible defaults, change only what you need.
Automatic pre-flight checks before the analysis runs.
Plain-language findings + quality verdict.
The statistical logic behind the result, explained in plain language.
Full APA-formatted report — tables, plots, metrics, and export.
Variables
Tell the analysis which columns to use. The variable picker automatically filters your dataset to show only columns compatible with the chosen method.
Step 1 — Variables
Numbers with meaningful magnitude — height, revenue, temperature, test scores.
e.g. age, price, duration, score
Labels with no inherent order — group names, product types, treatment conditions.
e.g. country, color, treatment_group
Categories with a meaningful order but unequal spacing.
e.g. S / M / L, 1–5 satisfaction, education level
Temporal data used to order observations or define a time axis.
e.g. order_date, timestamp, month_year
Settings
Configure the parameters for the chosen method. Every setting has an inline guide — hover the help icon to see what it does and the recommended default.
Step 2 — Settings
Validation
Before the analysis runs, Statistical Lab automatically checks your data — three categories, each returning pass, warning, or fail.
Step 3 — Validation
- Minimum sample size met for the chosen method
- Correct number and types of variables selected
- No empty columns or all-missing variables
- Normality where required — Shapiro-Wilk test + Q-Q plot
- Homogeneity of variance across groups — Levene's test
- Independence of observations
- Missing values detected — choose to drop, impute, or proceed
- Outliers identified via IQR and Z-score methods
- Constant or near-constant columns flagged
When a check fails: you can still proceed — checks are never hard blockers — but Statistical Lab shows a recommended fix directly under the failing check.
Common issues
Batch Analysis
Batch Analysis runs many analyses on the same dataset in one pass — assign your variables once, select every method that fits, and get a single integrated interpretation instead of opening each analysis one by one.
Upload your data
Drop a CSV or Excel (.xlsx / .xls) file, or load a built-in example dataset.
Assign variables once
Set an optional target / dependent variable and a group variable. They are reused across every analysis that needs them.
Select analyses
Tick individual methods, or use “Select all feasible” to auto-pick every analysis your data supports.
Run & interpret together
Run the batch, then click Generate for an AI interpretation that integrates all results into one narrative.
When to use it: exploratory passes over a new dataset, or producing a full battery of results for a report. For a single focused test, run the analysis directly from its own page.
Understanding results
Once an analysis runs, the result is presented across three tabs of increasing depth — the same finding, framed for different audiences.
Select the columns your analysis will use.
Configure parameters — sensible defaults, change only what you need.
Automatic pre-flight checks before the analysis runs.
For: Anyone
Plain-language findings + quality verdict.
For: Curious readers
The statistical logic behind the result, explained in plain language.
For: Researchers
Full APA-formatted report — tables, plots, metrics, and export.
Summary
Plain-language findings. Answers "what did the analysis find?" without statistical jargon.
Step 4 — Summary
Headline finding
One or two sentences stating what the analysis found in plain English — no jargon.
Key Findings panel
Bulleted highlights: key metrics, effect direction, group sizes, and anything notable about the data pattern.
Performance Assessment
A paragraph contextualizing the result — whether the model or test performed well, what the caveats are, and what the finding means in practice.
Quality Dimensions
A grid showing each quality dimension evaluated for this specific method, each with its own status badge. Dimensions vary by analysis.
Overall verdict chip
A single combined verdict at the bottom — e.g. "Moderate · 96.0%". This combines all dimension scores, not just the main metric.
"Why This Result?" button
Navigates directly to the Reasoning tab for a full dimension-by-dimension explanation.
Quality dimension badges
Reasoning
Answers "why did the result come out this way?" — translates statistical logic into plain language. Open when the verdict is Moderate or lower.
Step 5 — Reasoning
What the result means
Plain-language interpretation of the main finding — why the numbers came out the way they did given your data.
How reliable is this result?
Explains stability, variance, and consistency. For cross-validation: fold-to-fold spread and CV standard deviation.
What to consider next
Practical guidance: what the result supports, what it doesn't prove, and what follow-up steps would strengthen confidence.
Method-specific caveats
Warnings specific to the method — e.g. data leakage risks for cross-validation, multicollinearity for regression.
Diagnostic Guide
A reference table showing numeric thresholds for each dimension — what score range counts as Strong vs Moderate.
Bottom Line
A one-paragraph summary of the overall verdict with the key numbers — safe to use as a standalone takeaway.
Statistics
The full technical record — APA-formatted, with all numbers, tables, plots, and export options. Use when you need to publish, present, or hand off.
Step 6 — Statistics
Metric summary cards
Key numbers at a glance — the primary metric, spread/error, min/max, and sample size. Varies by method.
APA-style writeup
A publication-ready paragraph with all required statistics. Safe to copy directly into a manuscript.
Key Insights
Bulleted technical highlights — the most important numbers and what they imply, written for a statistically literate reader.
Visualizations
Charts specific to the method — Q-Q plots, residual plots, ROC curves, fold score charts, dendrograms, scree plots, and more.
Detailed tables
Coefficient tables, ANOVA tables, fold-by-fold results, cluster profiles, fit indices — whatever is standard for the chosen method.
Model parameters
A record of every setting used to run the analysis — so the result is fully reproducible.
Continue your analysis
Recommended next steps: Go Deeper, Consider Instead, and Related analyses tailored to your result.
Export menu
Download buttons for CSV, PNG, Word document, and reproducible Python / R code.
How to use all three tabs
Read the Summary.
Note the headline finding and check the overall verdict chip at the bottom.
If the verdict is Strong — you can stop here.
The Summary is self-contained. Report the headline finding with confidence.
If the verdict is Moderate or Weak — open Reasoning.
The Reasoning tab shows exactly which quality dimension is dragging the score down and what to do about it.
If you need to publish or present — open Statistics.
Copy the APA writeup, download the charts, or export the reproducible code.
Use "Continue your analysis" to decide what's next.
Go Deeper, Consider Instead, or Related analyses are suggested based on your result.
Common questions
Exporting & AI Chat
Take your analysis out of Statistical Lab — as data, an image, a Word document, or reproducible code. Or ask the AI Chat to help you interpret and communicate the result.
Export button location
CSV Spreadsheet
Further analysis in Excel, Google Sheets, or R / Python
Your original dataset enriched with the analysis output — cluster labels, predicted classes, residuals, or any other column the method adds.
- All original columns from your dataset
- New columns added by the analysis (e.g. predicted class, cluster label)
- One row per original observation
You want to keep working with the augmented data in another tool.
PNG Image
Slides, dashboards, quick sharing
A high-resolution screenshot of the entire Statistics tab — APA writeup, tables, and diagnostic plots captured as a single image.
- The APA-style writeup paragraph
- All result tables
- All diagnostic plots
- Captured at 2× resolution for sharpness on Retina displays
You need a single image to drop into a presentation, email, or chat message.
Word Document (.docx)
Papers, formal reports, manuscripts
A fully editable Word file with the APA writeup, formatted tables, and embedded plots — a complete first draft ready for editing.
- APA-formatted text paragraph — copy-paste ready
- Tables formatted with Word styles (editable)
- Embedded plot images
You are writing a paper, thesis, or formal report and want to start from a complete draft.
Python or R Code
Reproducibility, peer review, custom modification
The exact script used to produce the result — runnable in your own environment with the same parameters you set in the UI.
- Library imports and setup
- The analysis call with your exact parameter values
- Both Python (.py) and R (.R) versions available
You need to reproduce the analysis offline, share it with a collaborator, or extend it beyond what the UI exposes.
AI Chat
AI Chat — ask anything about your result
Once an analysis runs, a circular toggle button appears at the bottom of the Statistics tab. Click it to open the AI Chat panel — the AI already knows which analysis was run and what the result was, so you can ask follow-up questions without re-explaining the context.
Explain results in plain language
Translates statistical output into plain sentences — useful for stakeholders who don't read p-values.
Interpret key statistics
Ask about any number in the result — effect sizes, confidence intervals, fit indices — and get a clear explanation.
Draft an APA summary
Request a publication-style paragraph based on the result, ready to copy into a manuscript.
Suggest next steps
Ask "so what?" — what the result implies in practice and what you should consider doing next.
Which format do I need?
Continue analyzing the data in Excel or Sheets
CSVDrop a snapshot into a slide deck
PNGWrite a research paper or thesis
Word DocumentReproduce the result on your own machine
CodeHand the analysis off for peer review
Code + WordCaveats & tips
R code is independently written
Calculations on the platform run in Python. The R script is a faithful re-implementation, but small numerical differences can occur due to library differences. Verify before submission.
Word documents are starting drafts
The exported .docx is a complete first draft. Match the language to your journal's style guide before submitting.
PNGs are exported at 2× resolution
Images look sharp in slides and on Retina displays. If you need a vector format, use the Word document and extract the embedded plots.
Common issues
Guide & Terminology
Two reference tools you can open from any analysis page without losing your place — the Analysis Guide for the method you're using, and the Statistical Glossary for any term you don't recognize.
Analysis Guide
"Analysis Guide" button — top right of any analysis page
A method-specific reference explaining what the algorithm does, when to use it, what assumptions it makes, and what every parameter controls. Changes based on the analysis you're on.
Statistical Glossary
"?" button — top right of any analysis page
A searchable dictionary of statistical terms — p-values, effect sizes, confidence intervals, and method-specific vocabulary. Same on every page.
Guide & Terminology button location
Example — Cross-Validation Analysis Guide
What's inside every Analysis Guide
What is [Method]?
A plain-language explanation of what the algorithm does conceptually — how it works, what problem it solves, and how it differs from similar methods.
e.g. "Cross-validation is a technique for estimating generalization performance — how well a model will perform on unseen data."
Why use it?
The case for choosing this method over alternatives — key advantages, typical use cases, and what kind of data or research question it suits best.
e.g. "CV uses all data for both training and testing, reports mean AND std, and detects overfitting."
Methods / Variants
Where applicable, a breakdown of the variants available in the UI — what each one does, when to choose it, and practical tradeoffs.
e.g. K-Fold, Stratified K-Fold, Repeated K-Fold — each with when to use and what the tradeoff is.
Key insight
One or two sentences that capture the most important thing to understand about interpreting this method's output.
e.g. "The CV mean summarizes performance; the CV std measures stability — a lower mean but lower std may be more reliable."
Parameter reference
Every setting available in Step 2, with a plain-language description, the recommended default, and when you'd change it.
e.g. n_folds: "5 is a good default — 10 folds give lower bias but are slower and higher variance."
Statistical Glossary
Sample glossary terms
Common issues