0

I asked another question earlier and I think I miswrote that question.

My older question is here Can I merge multiple linear regressions into one regression?

In the older question I was talking about splitting dataset into multiple subsets and putting them in 4 different models. and I got the answer for that one.

My question this time is a bit different

As my dataset is consists of multiple features, Like this

Y     X1      X2      X3      X4
738   83      29      74      44
849   47      27      84      37
820   16      82      83      64
.
.
.

This dataset has 48,000 records

I have split this dataset into 4 different models So each model has 48,000 records with different features

enter image description here

Model 1 equation was $a_1 X_1 + b_1$

Model 1 equation was $a_2 X_2 + b_2$

Model 1 equation was $a_3 X_3 + b_3$

Model 1 equation was $a_4 X_4 + b_4$

Now I want to compile all 4 models into 1 model $Y_* = a_* X_{1,2,3,4} + b_*$

How can I achieve that?

asmgx
  • 239
  • 2
  • 9
  • 2
    Have you tried doing a multiple linear regression? – Demetri Pananos Oct 13 '21 at 04:16
  • Yes this is what I did. I did multiple linear regressions, each model was taking a different subset of features, but how to combine them all together in the end? – asmgx Oct 13 '21 at 11:00
  • That's not what a multiple linear regression model is. You are supposed to put all features in one model, not doing a separate model for each feature. – user2974951 Oct 13 '21 at 11:10
  • @user2974951 if i pulled all features in one model then it is not a multiple linear regression! it is a single one.. am I getting something wrong here?! – asmgx Oct 13 '21 at 12:27
  • Yes, you got it wrong. Multiple refers to multiple variables into one model, not to multiple separate models. – user2974951 Oct 13 '21 at 12:32
  • @user2974951 This is the point of my question, to split multi-variable model into multi-models. is this in any way possible? – asmgx Oct 13 '21 at 12:58
  • You can split the models, but you cannot join them, it is not that simple. I don't know why you are splitting them in the first place, but this is almost surely wrong. – user2974951 Oct 13 '21 at 13:02
  • @user2974951 I have reasons for splitting them, I know that combining algorithms in split on the row based is not possible. but not sure of the models split by the columns – asmgx Oct 13 '21 at 13:38

0 Answers0