4
$\begingroup$

Thanks in advance for the help.

I am writing a paper and for the life of me can't remember the proper term for a model that works as follows.

rawData -> model1 -> outputModel1 -> model2 -> outputModel2

More specifically, I have created a linear regression model whose output I pass to a second linear regression model. One could imagine N models strung together like this.

I keep wanting to call this a hierarchical or multi-level model, but I know that isn't right: these types of models are synonymous with mixed models. I know that there is a proper name for this since I remember discussing it in a class that took awhile back. Does anyone know what this is called?

$\endgroup$
  • $\begingroup$ The word ``cascaded'' comes to mind... $\endgroup$ – mcvz Jan 29 '15 at 7:17
  • $\begingroup$ Hierarchical and multilevel are certainly "reserved" for mixed models so using them for this kind of model would certainly be confusing. $\endgroup$ – Tim Jan 29 '15 at 7:18
  • 1
    $\begingroup$ "Hierarchical models" is rather generic, so I do not see why it would not apply. In Econometrics, you also have "simultaneous equation models" where outputs of some equations are input of others. $\endgroup$ – Xi'an Jan 29 '15 at 11:29
  • $\begingroup$ What about just "two-stage" or "two-step" estimation? $\endgroup$ – shadowtalker Jan 30 '15 at 5:53
4
$\begingroup$

The process is known as Cascaded classification/regression or Multi-stage classification/regression. It is a type of ensemble learning with some differences. You can find more in Wikipedia.

$\endgroup$
2
$\begingroup$

I am not sure, if there exists at all a "proper name" for models that you describe, but I would call such models chain models, similarly to the chain procedures, introduced in the following paper: http://www.multxpert.com/doc/md2011.pdf. This term IMHO better reflects the nature of this type of models and, at the same time, prevents potential confusion with hierarchical and multi-level ones.

$\endgroup$

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.