How to use the correlation matrix tool
Use this tool to calculate pairwise associations between selected numeric columns. It displays the result as a labelled matrix and heatmap, but it does not fit predictive models or alter your source values.
Upload a supported data file
Drop a .csv, .json, .xlsx, .xls, .tsv, or .txt file into the upload area, or click browse files. The file input accepts all of these extensions, but text files are read with the tool's comma-delimited parser. The first row becomes the column headers.
For JSON, provide an array of records or an object containing an array of object records. Spreadsheet files are read from the first workbook sheet only.
Configure the calculation
Choose one of the available methods:
- Pearson (linear) measures linear association.
- Spearman (rank, monotonic) uses ranks to measure monotonic association.
- Kendall tau-b (rank, monotonic) compares the ordering of paired values and applies a tie correction.
Under missing values, choose Pairwise deletion to use valid rows separately for each pair, or Listwise deletion to remove rows missing a value in any selected column. Set Minimum pairs and Decimal places as needed. An invalid or too-small minimum is normalized to at least 2; an invalid value uses 3.
Select columns and calculate
The tool detects numeric columns automatically. A column is included when at least 60% of its non-missing values pass the numeric parser. The parser accepts plain numbers, comma-separated numbers, scientific notation, and percentages. Constant columns are detected but start unchecked.
- Filter the list if needed, then select at least two numeric columns.
- The matrix updates automatically whenever the file, selected columns, or calculation settings change.
Read the results
Each off-diagonal cell shows an association value for a selected pair. With Pairwise deletion, different cells can use different numbers of valid observations. Diagonal cells show the numeric count for that column. A constant column or a pair below Minimum pairs is shown as N/A.
Enable Show significance (p-value) to change the displayed cell detail to p-values. This also adds a separate p-values section to the CSV export. Correlation describes association; it does not establish causation.
Export CSV or JSON
After calculation, use Export CSV or Export JSON. CSV contains the matrix and, when Show significance (p-value) is enabled, a separate p-values section. JSON contains calculation metadata and pair records rather than a two-dimensional matrix object. Each pair record includes r, n, p, na, and reason.
Limits and troubleshooting
- Empty files show an error. Invalid JSON or JSON without a suitable record array cannot be processed.
- If no numeric columns are detected, the column list remains empty.
- Fewer than two selected columns prevents calculation.
- Listwise deletion can remove every row if no row has valid values in all selected columns.
- A result can show
N/Awhen a column is constant, a pair has too few valid observations, or the calculation is undefined.
Privacy and local processing
The selected file is processed locally by this tool in your browser.
Frequently Asked Questions
Does correlation imply causation?
No. A high correlation shows association between two columns, not that one causes the other.
What is the difference between Pearson, Spearman, and Kendall?
Pearson measures linear association. Spearman and Kendall use ranked values to measure monotonic association; Kendall tau-b also applies a tie correction.
How is missing data handled?
Pairwise deletion uses valid rows separately for each pair. Listwise deletion drops rows missing any selected column before computing the matrix.
Does this tool modify my data or build a predictive model?
No. It reads the file to calculate values for display and export; it does not rewrite source values or fit a predictive model.
Why does a pair show "N/A"?
One column may be constant, or the pair may have fewer valid observations than the Minimum pairs value.