I'm using generalized linear models and generalized nonlinear models in a modeling project, so I refer to the final predictor simply as the predictor. But I find that this is confusing and unclear, as you can also refer to the predictor variables as predictors. Is there another way to refer to the final predictor, or any other way to distinguish the final predictor and predictor variables?

  • $\begingroup$ I think @IsabellaGhement's answer (+1) addresses your concern quite well. Another word you could use is: prediction machine. $\endgroup$ – Jim Mar 18 '18 at 19:11

I think it's a bit odd to talk about the "final predictor". Usually, one talks about the "final model" and then elaborates on what predictors were included in that model (relative to the "initial model").

You may find the post available at https://www.theanalysisfactor.com/whats-in-a-name-moderation-and-interaction-independent-and-predictor-variables/ interesting. The post talks about the terminology "predictor variable" and mentions that there are many subtypes of predictor variables.

If your models are used for prediction, then the terminology "predictor variable" certainly makes sense. If they are used for explanation (i.e., to uncover and quantify associations), perhaps you could use the terminology "explanatory variable" instead of "predictor variable". If they are used for both explanation and prediction, you could revert back to the "predictor variable" terminology.

In any event, if you make it clear what predictor/explanatory variables you started out with in your "initial model" and what predictor/explanatory variables you were left with in your "final model", there should be no confusion.

  • $\begingroup$ Right, actually that's not what I meant by 'final predictor', but it's probably poor expression on my part. I'm not starting with a full model then reducing it to get a final model, I'm simply refering to the predictor that you get from applying a function to predictor variables $\endgroup$ – liyuan Mar 18 '18 at 14:51
  • $\begingroup$ I see...Then maybe just call it the "transformed predictor" and explain what transformation you used. For instance, if your original predictor was "age" and you log-transformed it prior to including it in your model, then call it "log-transformed age". If the transformation is more complex, just call it the "transformed predictor". $\endgroup$ – Isabella Ghement Mar 18 '18 at 15:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.