Hi,
Dear forum,
When doing a polynomial regression with =LINEST for three independent variables, one should use an array after the input-variables to indicate the degree of the polynomial intended for that variable.
What I would like to do is making a regression for three variables like this:
The intended model Y = A + (B*X1 + C*X2)(1 + D*X3)
How should I find the coefficients?
LINEST() requires that your model be in a form Y=a*f1(x1,x2,x3)+b*f2(x1,x2,x3)+c*f3(x1,x2,x3)+... -- in other words, it must be a linear function where the parameters are linear. In this case, when you expand the multinomial, you get B*D and C*D terms which are not truly linear parameters.
It's possible that, if you spend enough time with the algebra that you can rearrange that into a linear function. However, I would suggest that it might be easiest to approach this using a non-;linear regression algorithm using Solver, rather than trying to coax this into something linear that fits into LINEST(). Would that work for you?
Originally Posted by shg
You could regress y = m1 x1 + m2 x2 + m3 x1 x3 + m4 x2 x3 + b using LINEST.
Last edited by shg; 04-06-2017 at 01:46 PM.
Entia non sunt multiplicanda sine necessitate
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
LinkBack URL
About LinkBacks
Register To Reply
Bookmarks