Analyses

Principal component analysis

Compress correlated variables into a smaller set of components.

Edit on GitHub

When to use it

Use PCA when you have several correlated scores and want a few components that keep as much of the variance as possible. It is a compression of the variables, not a test that a particular factor model is true. For a stated factor model, use confirmatory factor analysis.

Assumptions

Variables are numeric. Tensr standardizes them first, so the analysis is on the correlation matrix, not the covariance matrix. Rows with any missing value are dropped. There is no KMO or Bartlett test on this request.

Running it in Tensr

Analyze → Dimension Reduction → PCA. In chat: “PCA of these six items, two components.”
Select at least two columns. Leave components empty to keep every component that fits, or set a number from 1 to 50.

Options

Prop

Type

The wizard may send an empty string for the component count. That has to become null before it reaches the API. A blank field means “let the server decide.”

Reading the output

Six items from that same scale. The first component should take a clear share and leave a real second component.

ComponentEigenvalueVariance %Cumulative %
PC12.28637.782%37.782%
PC21.02416.923%54.705%
VariablePC1PC2
item10.378-0.316
item20.426-0.429
item30.448-0.32
item40.4490.231
item50.4140.291
item60.3210.69

PCA on 6 variables, n = 120. PCA on 6 variables, n = 120. Metrics: Components = 2; PC1 variance = 37.782%.

Reporting (APA 7)

PCA on 6 variables, n = 120. Report the estimate in the table. This procedure is not summarised by one p-value.

Coming from SPSS

Analyze → Dimension Reduction → Factor, Extraction: Principal components. SPSS can also print KMO and Bartlett’s test. This request does not.