What happens after you load a file
After a file loads, the workspace lists the numeric columns it detected. Choose the calculation settings, keep the columns you want, and the matrix updates when at least two columns are selected. Each off-diagonal cell contains a coefficient or N/A; each diagonal cell shows the numeric count for that column.
- Drop a CSV, JSON, or spreadsheet file into the upload area, or click
browse files. - Choose a
Method,Missing valuesmode,Minimum pairs, andDecimal places. - Filter the detected columns with
Filter columns, then select at least two numeric columns. - Review the matrix, then use
Export CSVorExport JSON.
What each correlation method measures
Pearson (linear) uses the numeric values to measure linear association. Spearman (rank, monotonic) and Kendall tau-b (rank, monotonic) compare ordering. Kendall tau-b accounts for tied values. The heatmap colors negative coefficients red, positive coefficients blue, and values near zero gray.
How columns enter the numeric list
A column enters the list when at least 60% of its non-missing values pass the numeric parser. The parser accepts plain numbers, commas, scientific notation, and percentages. It converts a percentage to a fractional value, so 25% becomes 0.25. Blank strings and tokens such as NA, N/A, null, NaN, and - count as missing.
CSV, TSV, and TXT files use commas as delimiters, with the first row used as headers. JSON must contain an array of records, either at the top level or inside an object. XLSX and XLS files use the first workbook sheet.
How missing values affect the matrix
Pairwise deletion finds valid numeric rows separately for each pair, so cells can have different observation counts. Listwise deletion keeps rows with valid values in every selected column. If no row remains, the tool does not build the matrix.
Constant columns start unchecked and produce N/A when selected. A pair also shows N/A when it has fewer valid observations than Minimum pairs or its calculation is undefined. The tool treats Minimum pairs as at least 2; an invalid value falls back to 3.
sales,visits
12,100
18,120
selected columns: sales, visits
off-diagonal cell: N/A (below minimum pair count)The sample has two valid rows, so the default minimum of 3 leaves the pair as N/A. Add a valid row or change Minimum pairs to 2.
Export CSV writes the labelled matrix. When Show significance (p-value) is enabled, the file adds a separate p-values section. Export JSON includes the selected columns, calculation settings, and pair records with r, n, p, na, and reason. The source values are not rewritten, and the exported note describes correlation as association rather than causation.
What to check before exporting
Check the selected columns and the warning below the column controls. Cell details include the pair count, and the heatmap shows the coefficient's sign. If a pair is marked N/A, inspect its observation count and the selected Minimum pairs.
For data with blank cells, compare Pairwise deletion and Listwise deletion before saving the result. The two settings can leave different rows available for calculation.
When another dataset tool fits better
Use Dataset Analyze Tools for several dataset checks in one place. The Dataset Profiler is better for column types, row and column counts, null and blank rates, unique and duplicate counts, and column statistics.
Choose the Fuzzy Duplicate Finder to review likely duplicate records and choose which record to keep or merge. Choose the Time-Series Anomaly Detector when timestamps, gaps, duplicates, or rolling anomaly thresholds are the focus.
Frequently asked questions
What file can I use to calculate a correlation matrix?
Use CSV, TSV, TXT, JSON, XLSX, or XLS files. CSV-style files use the first row as headers, JSON needs an array of records, and spreadsheets use the first workbook sheet.
How many columns do I need for a correlation matrix?
You need at least two selected numeric columns. The matrix stays unavailable until two columns are selected.
Should I use pairwise or listwise deletion?
Use Pairwise deletion when each pair should use its own valid rows. Use Listwise deletion when every selected column must have a valid value in the same row.
Why does my correlation matrix show N/A?
N/A means a column is constant, a pair has fewer valid observations than Minimum pairs, or the selected calculation is undefined. Check the warning below the column controls and the pair count in the matrix cell.
Does the correlation matrix change my uploaded file?
No. The selected file stays in the browser while the matrix is computed, and the tool does not rewrite the source values.