Suppose If I have a dataset in which few variables are linear and few are neutral (like JR scale in MR) and I want to see the drivers of a linearly continuous variables on the basis of both the set of variables. we can run a linear regression for that purpose but here to include neutral variables as exploratory vars will not be statistically valid, so in order to perform a universal kind of analysis which alternative regression technique would be best?
& Yes I don't want to go for SEM.
#bayesian #SVM #BayesianBeliefNetwork #SEM
I propose you take a look at zunzun / pyeq3 project, aka multi formulated regression.
https://github.com/zunzun/zunzunsite3
https://github.com/zunzun/pyeq3
I find their work fascinating. :)
Emerging of machine learning gives us a possibility to predict an outcome at the basis of explanatory variables. The good thing is that machine learning needs neither any assumptions nor shape of function binding the variables. The program just "learns" what is the structure of the variables. I recommend especially artificial neural networks that is a kind of "regression". You will find plenty of information in the Internet about neural networks. You may run neural networks e.g. in STATISTICA (by StatSoft).
Good luck!
Thanks Michal S. Karbownik.
Even I was also wondering to explore machine learning techniques.
ANN would also be a good technique, I'm working on ANN too.
I have been working on Bayesian Belief Networks using BayesiaLab and R.
They both have a flexibility
Thanks a lot Karbownik for adding your answer .
Rgds,
Abhay
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression