topic Model screening vs Model Comparison in Discussions

Model screening vs Model Comparison

sofiasousa24 — Thu, 01 Feb 2024 12:40:29 GMT

I aggregated data from various DoEs into a single data matrix table. Now, I'm trying to select the best predictive model to fit my data.

Which jmp platform is the best? Should I use model comparison or model screening? I'm not sure which platform suits my purpose and my data. Is there literature or a webinar I can use to understand the main differences between the two platforms?

Thank you so much.

Re: Model screening vs Model Comparison

Mark_Bailey — Thu, 01 Feb 2024 14:22:15 GMT

What is your purpose? Are you searching for an explanatory model or a predictive model?

Re: Model screening vs Model Comparison

statman — Thu, 01 Feb 2024 14:47:26 GMT

Watch this:

https://www.jmp.com/en_us/events/ondemand/analytically-speaking/to-explain-or-predict-that-is-the-question.html

Re: Model screening vs Model Comparison

sofiasousa24 — Sun, 04 Feb 2024 10:40:28 GMT

Hi @Mark_Bailey

I've done some experimental designs in the context of formulation optimisation (custom designs and mixture designs) with the main objective of understanding the impact of each of the variables on the response - drug release at different time points. Now, I'd like to build a predictive model of drug release based on a general data matrix that aggregates all the different DoEs. Does that make sense? For that, I’m thinking of conducting a model screening to choose the best predictive model.

Additionally, in the model screening platform, the model report appears for each continuous response (Y). Is it possible to model all the responses simultaneously? I have 7 continuous responses so, how can I choose the best predictive model for whole matrix data?

I'm sorry if I'm not making myself clear.

Thank you so much for your help!

Re: Model screening vs Model Comparison

peng_liu — Tue, 06 Feb 2024 03:34:08 GMT

Between model comparison or model screening , I believe that Model Screening is what you need. Because Model Comparison requires you to build models first and generate prediction columns from those models. It sounds like that you don't know any specific models that you favor yet. Then Model Screening provides you a list of popular predictive models.

You can enter multiple Y's as you see. In general, your decision still needs to be made for one response at a time. The software is designed to help you to find the best model for every Y. If your overall criteria is different from "finding best model for individual Y", then you need to elaborate.