I was wondering how PRPC handles the predictors in an ADM model. I realize that the more predictors listed the more of a load it would be on the system, but what if some of the listed predictors do not have any value (null) when the model is called? for example lets say one lists over 1k of properties as predictors but each time that the model runs say only 100 of them actually have value is the system then taxed for the 1k listed or would it behave as if only 100 were listed?
I was also wonder if someone could give an example of how to utilize the Parameters in an Adaptive model, or the benefits of using Parameters vs actual properties.
Hahaha, I knew i might have entered chartered territory here with number of predictors, and judging by the lack of replies i might have been right. Here is some feedback from my own experience.
1. Rule Limitations:
a. It seems if Predictors have null values it fails during run-time, not sure if this is true but i was getting this once i entered a parameter as a predictor and forgot to assign a value to it.
b. Once you go over 680 listed predictors the rule will not compile, hard limit here from java standpoint.
As far as the parameter question was concerned i guess i get the value in that you can set it up as mapping exercise via Data-Transforms but my initial confusion was related to thinking that the predictors would have to support a more stable relationship between the model and the clipboard objects. Meaning that if you pointed it to a whole different property later it would actually interfere with the model's performance. But i guess you could point it to a different location where that "original" property exists in cases where it might change in the future.
Feel free to correct me on any of the above, its mostly based on my quick test and limited experience, although i would not recommend you try to add the 680+ predictors lol.