I HAVE A VARIABLE' Y', DEPENDENT ON THREE PARAMETERSX1 , X2 , X3


I WANT TO DEVELOP AN EXPRESSION OF the form
y=(ax1^3/(bx2+cx3))+bx1^2+cx1^3

where symbol^ refers to index.
IS IT POSSIBLE IN EXCEL

Kind of, kind of not.

It's not a standard regression by any means, so one way would be to use
SOLVER to minise the squared error with a lot of iterations. Remember to use
a pretty good initial estimate, though. It's not really an analysis, since
it's just as estimating thing. Any diagnostics will need to be done manually.

Although, I must say, that's a very unusual model that I haven't seen in my
many years of financial modelling...

Another way to do it would be to use a log transform. That is, log transform
both sides to make it easier to set up the columns then use the usual excel
regression tool. Again, this isn't quite good either.

It's a very unusual model, if I can say so again.

It's very non-linear is every sense of the word. If solver doesn't hack it,
then the only other way I can think of is to set up an Artificial Neural
Network routine, which can be done in excel.

"SANJAY" wrote:

I HAVE A VARIABLE' Y', DEPENDENT ON THREE PARAMETERSX1 , X2 , X3


I WANT TO DEVELOP AN EXPRESSION OF the form
y=(ax1^3/(bx2+cx3))+bx1^2+cx1^3

where symbol^ refers to index.
IS IT POSSIBLE IN EXCEL