Analysis results
Statistical summary
Machine learning insights
- Insights appear after analysis.
Correlation matrix
Multicollinearity report
Formula used
r = Σ[(xᵢ − x̄)(yᵢ − ȳ)] / √(Σ(xᵢ − x̄)² Σ(yᵢ − ȳ)²)
z = (x − μ) / σ
ŷ = b₀ + b₁x
R² = 1 − Σ(yᵢ − ŷᵢ)² / Σ(yᵢ − ȳ)²
VIFⱼ = 1 / (1 − Rⱼ²)
Lower = Q₁ − k·IQR, Upper = Q₃ + k·IQR
How to use
- Paste data or load an example dataset.
- Parse the data and select numerical variables.
- Choose grouping, preprocessing, plotting, and outlier settings.
- Generate the matrix and review correlations, statistics, and insights.
- Export the chart, cleaned data, reports, or configuration.
Example data format
| sepal_length | sepal_width | petal_length | petal_width | species |
|---|---|---|---|---|
| 5.1 | 3.5 | 1.4 | 0.2 | setosa |
| 6.4 | 3.2 | 4.5 | 1.5 | versicolor |
| 6.3 | 3.3 | 6.0 | 2.5 | virginica |
Interpretation guide
Upward point patterns suggest positive relationships. Downward patterns suggest negative relationships. Curves may indicate nonlinear effects.
Tight groups can reveal class separation. Isolated points may be outliers. Dense overlap can hide useful structure.
High correlations can indicate redundant predictors. Visual relationships do not prove causation. Confirm findings using validation and domain knowledge.
Limitations
Large matrices become difficult to read. Sampling may hide rare observations. Missing data choices can change correlations.
Correlation measures association, not causation. Outlier rules are screening tools only. Transformations alter the displayed scale.
VIF estimates can be unstable with limited rows. Use enough observations. Verify important findings with formal modeling.
Frequently asked questions
What is a scatterplot matrix?
It displays pairwise scatterplots for several numerical variables. Diagonal cells summarize each variable. This supports rapid exploratory analysis.
How many features should I select?
Four to six features are usually readable. More variables create smaller cells. Select only features relevant to the analysis.
Which correlation method should I use?
Pearson measures linear association. Spearman measures monotonic rank association. Kendall is robust for smaller samples.
Why are some rows excluded?
Rows may contain missing or invalid values. Filters can also remove observations. The results report displays the final count.
What does a high VIF mean?
High VIF suggests predictor redundancy. Coefficients may become unstable. Consider removing or combining related variables.
Can categorical variables be plotted?
Categorical variables can group or color observations. Matrix axes require numerical values. Encode categories only when mathematically justified.
How are outliers detected?
IQR uses quartile-based fences. Z-score uses standardized distance. Both methods require careful interpretation.
Does the calculator upload my data?
Analysis runs inside your browser. This page sends no dataset to PHP. External libraries still load from their providers.
Can I save my settings?
Use the configuration export button. It downloads a JSON settings file. Load that file to restore options.