FDE: Using the model and FPCs to assess FPCs of a new curve

martindemel

Hi,

when using FDE to generate FPCs we can use the profiler to generate almost every shape we wat and see what FCs are required to do this. However, as soon we have a new curve and want to know what FPCs belong to this curve we would need to fit all curves again and the previous found model for the other curves got overwritten (also with new FPCs). I understand why.

BUT, is there any way to have FPCs build on previous curves and then just predict FPCs based on this model for the new curve? For interpretability it would change the world as now the FPCs change as soon you build a new fit and all the interpretations previously done are not comparable anymore.

 

Any ideas?

@chris_gotwalt1 , 

 

chris_gotwalt1

It is a little awkward to do this currently. The way to do that currently involves

(1) Fitting the model to the original curves and saving the script that fits the model someplace.

(2) Appending the new curves to the original table, and labeling them validation rows in the Validation column.

(3) Run the script and when you save summaries you will have the FPC scores for the new functions.

I admit this is neither obvious nor ideal in terms of steps. The "scoring problem" has proven difficult to implement in part because of the built in data transformation steps. So, we have chosen easier wins. At some point we need to find a way to make this easier and more transparent.