PCA Quick Calculator

Enter Dataset and Options

Paste rows of numeric observations. Separate values with commas, tabs, semicolons, or spaces.

Variable Labels

Components to Retain

Scale Variables First?

Decimal Places

Observation Matrix

Tip: Each row is one observation. PCA needs at least three rows and two variables.

Example Data Table

Observation	Revenue	Margin	Satisfaction	Returns
Obs 1	120	28	82	4
Obs 2	135	31	88	3
Obs 3	150	35	90	2
Obs 4	98	22	70	6
Obs 5	110	25	75	5
Obs 6	160	36	92	2

Formula Used

1. Center or standardize variables: For each variable, subtract the mean. If scaling is enabled, divide by the sample standard deviation as well.

2. Build the analysis matrix: The calculator forms a covariance matrix for centered data or a correlation matrix for standardized data.

3. Solve the eigensystem: PCA finds eigenvalues and eigenvectors of the matrix. Eigenvalues measure captured variance. Eigenvectors define component directions.

4. Compute explained variance: Explained percentage for component k is λ_k / Σλ multiplied by 100.

5. Compute loadings and scores: Loadings equal eigenvector coefficients scaled by the square root of eigenvalues. Scores equal the processed data matrix multiplied by retained eigenvectors.

How to Use This Calculator

Enter comma-separated variable labels that match the number of columns in your dataset.
Paste your observation matrix into the textarea, using one row per observation.
Choose how many components you want to retain in the final result.
Select whether variables should be standardized before PCA starts.
Press Calculate PCA to display the summary above the form.
Review explained variance, loadings, matrix values, and component scores.
Use the CSV or PDF buttons to export the generated report.

FAQs

1. What does this PCA calculator do?

It reduces many correlated variables into fewer principal components, then reports eigenvalues, explained variance, loadings, processed matrix values, and score previews.

2. When should I standardize variables?

Standardize when variables use different units or ranges. This prevents large-scale variables from dominating the covariance structure and often makes comparison fairer.

3. What is the difference between loadings and scores?

Loadings show how strongly each original variable contributes to a component. Scores show where each observation falls on those new components.

4. How many components should I keep?

Retain enough components to capture useful variance while keeping interpretation simple. Common checks include cumulative variance, scree shape, and the Kaiser rule.

5. Why do I get an error for zero variance?

A variable with no variation cannot be standardized and adds no PCA information. Remove constant columns before running the analysis again.

6. Does PCA work with very small samples?

It can run on small samples, but stability and interpretation improve with more observations. Extremely small datasets may produce fragile components.

7. Can I paste space-separated data instead of commas?

Yes. The parser accepts commas, semicolons, tabs, and spaces, as long as every row contains the same number of numeric values.

8. What do larger absolute loadings mean?

Larger absolute loadings indicate a stronger relationship between a variable and a component. Sign direction matters, but magnitude usually drives interpretation.