Applying functional PCA outcome to the new observed functional data

[#1]

I have functional curves from id 1 to id 80. I could obtain the functional PCA analysis result from JMP functional data explorer. Now, I want to obtain corresponding fPCA scores for new observed curves, say id 81. How can I do this?

[#2 · Accepted solution]

The current way to do this is to create a validation column (if you don't have one already) so that the new observations are considered validation data. When you re-launch FDE, the models will be re-fit to the original training data set holding out the validation rows.

The FPC scores will be calculated for the validation data, and you can Save Summaries to place them into a data table.

[#3 · Accepted solution]

It is possible, but there are a few steps.

1. Keep the original function data (curves 1 to 80).
2. Save a script for the original analysis. You might want to remove unnecessary models and just keep the one for the FPCs you want to use in the future.
3. Make a Validation column and fill it with 0 for the rows with the original functions.
4. Add rows with the new function (curve 81) and assign Validation of 1.
5. Run the script that was used to fit the original functional model but make sure that the Validation data column is cast in the validation analysis role. FDE will have to fit the model all over again to the training data, but then it will simultaneously score the rows in the validation set.
6. Save Summaries to extract the FPC’s.