Non Contiguous Columns / Index & Linest / Solver

  • Thread starter Thread starter Marston
  • Start date Start date
M

Marston

I have a series of "n" columns of equal length.

Each of these columns of data are candidates for a multiple variable
regression.
One column represents the dependent data, the other columns the
independent data.
Over time which columns of data in collection create the best
regression varies in time.

My definition of best columns for the regression is that for each
column of data:

abs(Index(linest(Ax:Ay, non-contigous columns,0,true),n,1)/
Index(linest(Ax:Ay,non-contigous columns,0,true),n,2))>2

Index(linest(Ax:Ay,non-contigous columns,0,true),1,3) >
Index(linest(Ax:Ay, non-contigous columns (less 1),0,true),1,3) >0

There's a little I'm leaving out like the fact that linest orders the
results in the array backwards from the order the columns are ordered
from the second range of values - but I know how to handle this. Also
the second condition is a bit of a simplification - but again, that
something I understand.

What I don't know is how to get around the issue that linest requires
that columns be contiguous in order to conduct its analysis.

Is there any function that could do something like new array =
(A1:A100, C1:D100, Z1:Z100) and it lines up the columns as if
A,C,D,and Z were contigous?

Thanks in advance

Marston
 
one other comment - the number of non-contiguous columns I'd like to
put together is not constant. It could be from 2-16
 
Back
Top