Can I calculate the scores of new test subjects from an existing PCA?

Yes, Engelbert Buxbaum, you can calculate the component scores for new test subjects using the existing loading matrix from your principal component analysis (PCA). Here’s how:

Standardize the New Data: Use the means and standard deviations from your original dataset to standardize the new data.

Project onto Principal Components: Multiply the standardized new data by the loading matrix from the original PCA to get the component scores.

To streamline this process, you might find OriginPro software particularly useful. It offers powerful tools for PCA and can easily handle the standardization and projection steps for new data.

Richroy Richard

Yes, you can use the existing loading matrix from your principal component analysis (PCA) to calculate the component scores for new test subjects. Here’s how you can do it:

Center the new data: Ensure the new test subjects' data is centered using the same means as the original dataset. If you originally standardized the data (i.e., scaled to have zero mean and unit variance), you should standardize the new data in the same way using the means and standard deviations of the original dataset.

Calculate component scores: Use the loading matrix from the original PCA to transform the new data. The component scores for the new subjects can be calculated by multiplying the centered new data matrix with the loading matrix (principal components).

Here is a step-by-step outline of the process:

Step-by-Step Process

Center (and potentially scale) the new data:Calculate the mean (and standard deviation, if scaling) of the original data. Subtract the mean (and divide by the standard deviation, if scaling) from the new data to center (and scale) it.

Calculate the component scores:Let XnewX_{\text{new}}Xnew be the matrix of centered new data. Let PPP be the loading matrix from the original PCA. The component scores TnewT_{\text{new}}Tnew for the new data can be obtained by: Tnew=Xnew⋅PT_{\text{new}} = X_{\text{new}} \cdot PTnew=Xnew⋅P

Example

Suppose you have the following:

μ\muμ (mean vector of original data)
σ\sigmaσ (standard deviation vector of original data, if you standardized)
PPP (loading matrix from PCA)

Let XnewX_{\text{new}}Xnew be the new data matrix.

Xnew, standardized=Xnew−μσX_{\text{new, standardized}} = \frac{X_{\text{new}} - \mu}{\sigma}Xnew, standardized=σXnew−μCenter the new data:Xnew, centered=Xnew−μX_{\text{new, centered}} = X_{\text{new}} - \muXnew, centered=Xnew−μIf standardized:

Tnew=Xnew, standardized⋅PT_{\text{new}} = X_{\text{new, standardized}} \cdot PTnew=Xnew, standardized⋅PCalculate the component scores:Tnew=Xnew, centered⋅PT_{\text{new}} = X_{\text{new, centered}} \cdot PTnew=Xnew, centered⋅PIf standardized:

This approach allows you to project the new subjects' data onto the principal components derived from the original dataset, thus enabling you to analyze the new subjects in the context of the existing PCA model.

How to compare two groups with only two measurements?

Enhanced Yellow Fluorescent Protein (Aequorea victoria) DNA sequence?

What kind of fluid could I use for a pitfall trap that also does not invalidate molecular testing?

What's the best way to measure growth rates in House sparrow chicks from day 2 to day 10?

Do we need to write protocol for a second study of a systematic review that has been registered and published?

How to increase a model accuracy ?

Do I need to do linear regression before running mediation analysis using PROCESS MACRO by Hayes ?

Sample dilution for quantitative analysis using ELISA?

Is There Any Feasible Method To Test Fluorescent Compounds Other Than Fluorescence Spectrometers ??

How to calculate average trend in maximum annual NDVI from MODIS?