PCA Rotation Tool Calculator

The table below shows a sample loading matrix you can paste.

This tool applies an orthomax rotation to an unrotated loading matrix L. The rotated loadings are computed as:

L_rot = L · T

The orthomax family uses a parameter γ:

• Quartimax: γ = 0
• Varimax: γ = 1
• Equamax: γ = p/2 (p = number of variables)

Communality for variable i is h²_i = Σⱼ L_rot(i,j)².

1. Paste your unrotated loading matrix.
2. Enable label options when your matrix includes names.
3. Choose a rotation method and optional Kaiser normalization.
4. Set output formatting, then rotate the loadings.
5. Review communalities, variance, and the rotation matrix.

Rotation aims and practical outcomes

A rotated loading matrix helps you see which variables define each component. This tool starts from unrotated PCA loadings and searches for a simple pattern: high loadings concentrated on fewer components and near-zero elsewhere. In applied studies, analysts often treat absolute loadings above 0.30 as meaningful and above 0.50 as strong. The highlight and cutoff controls let you apply those thresholds consistently across reports and teams. That improves interpretability without changing variance.

Choosing an orthogonal method

Varimax (γ=1) is the default because it maximizes variance of squared loadings within each component, usually yielding clear clusters. Quartimax (γ=0) emphasizes simplifying rows, so each variable loads strongly on one component, which can suit screening tasks. Equamax balances both by setting γ=p/2, where p is the number of variables. When you already know your preferred trade-off, Custom Orthomax lets you enter γ directly and compare results side by side in practice.

Normalization and convergence controls

Optional Kaiser normalization rescales each variable by its communalities during rotation. This prevents variables with large magnitude from dominating the criterion, especially when measurement scales differ. After optimization, the tool rescales back to original units, so your final loadings remain comparable. Use max iterations to allow difficult solutions to settle, and use tolerance to stop when updates become tiny. If convergence is slow, relax tolerance slightly rather than forcing thousands of iterations.

Reading the rotated table

After rotation, examine each variable’s largest absolute loading and check that it matches a coherent construct. The Communality column reports h², the sum of squared rotated loadings for that row; low values suggest the variable is poorly represented by the selected components. The Variance table lists each component’s sum of squares and percent of total rotated variance. With orthogonal rotation, the total sum of squares remains constant, though it redistributes across components.

Exporting and documenting results

To support review, export the rotated table as CSV for spreadsheets, or download a PDF for archiving and approvals. The results panel also includes a rotation matrix T, which documents the transformation from your original loadings. Saving T is useful when you need to reproduce scores or rotate new variables into the same solution. For transparent reporting, record the method, γ value, Kaiser setting, tolerance, and iteration count shown above the tables.

1) What input does the tool require?

Paste an unrotated loading matrix where rows are variables and columns are components. You may include variable names and component labels, or let the tool generate them automatically.

2) Does rotation change total explained variance?

For orthogonal rotations, the total sum of squared loadings is preserved. Rotation mainly redistributes variance across components to make the pattern easier to interpret.

3) When should I enable Kaiser normalization?

Enable it when variables have very different communalities or when a few high-magnitude rows dominate the solution. It can stabilize the optimization and produce more balanced simple structure.

4) How do I choose γ for Custom Orthomax?

Start with γ=1 to mimic Varimax, then test nearby values such as 0.5 or 1.5. Lower γ tends to simplify rows; higher γ tends to simplify columns.

5) Why are some cells blank in the rotated table?

If you set a suppression cutoff, the tool hides absolute loadings below that threshold to reduce clutter. Set cutoff to 0 to display all values.

6) Can this tool perform oblique rotations?

No. This calculator focuses on orthogonal rotations that keep components uncorrelated. If you need correlated factors, use an oblique method such as Promax in a dedicated factor analysis workflow.