Improve regression via Solver
Hi,
I'm using Excel Solver to run regressions on 63 factors and getting below par results.
Can anyone advise how I can beef up the resultant coefficients. ?
How can I determine factors that have a correlation ?
One problem which I have no power over is that some of the ranked variables that I am working with
may already be correlated.
However if I can separate any going forward it may well be beneficial.
Would adding one column at a time be advisable.
Any help on regression in general and in particular Excel's SOLVER would be most gratefully received.
Thanks
Ian
I'm using Excel Solver to run regressions on 63 factors and getting below par results.
Can anyone advise how I can beef up the resultant coefficients. ?
How can I determine factors that have a correlation ?
One problem which I have no power over is that some of the ranked variables that I am working with
may already be correlated.
However if I can separate any going forward it may well be beneficial.
Would adding one column at a time be advisable.
Any help on regression in general and in particular Excel's SOLVER would be most gratefully received.
Thanks
Ian
ASKER
Hi Jan
I have just built a model using the 63 factors and of those 29 produced positive coefficients.
Anyway, here is a description.
It is a horse race betting project involving 63 factors which can be culled via research.
There are 2869 races of which 1030 were selected by the model
Total rows in worksheet = 30546 The total races 2869 30546/2869 = 10.46 runners per race.
I did look at the site you recommended but being a beginner I would find it a bit beyond my ability.
I have just built a model using the 63 factors and of those 29 produced positive coefficients.
Anyway, here is a description.
It is a horse race betting project involving 63 factors which can be culled via research.
There are 2869 races of which 1030 were selected by the model
Total rows in worksheet = 30546 The total races 2869 30546/2869 = 10.46 runners per race.
I did look at the site you recommended but being a beginner I would find it a bit beyond my ability.
Sounds like you need to do multivariate linear regression analysis. Software like spss would aid with that.
ASKER
I do have Stata but unsure how to go about it. Solver seems an easy option
Stata seems to offer that feature, no idea how well it does it though.
ASKER
I've no idea how to use Stata.
ASKER
I am looking at Frontline who sell Solver licenses
ASKER
Hi Jan, if you are still about.
I have installed SPSS if you or anyone can advise how to use the 'multivariate solution'.
Thanks
I have installed SPSS if you or anyone can advise how to use the 'multivariate solution'.
Thanks
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
63 factors is a lot. How many measurements do you have? And are the measurements properly spaced out so all 63 factors can actually be determined from the data? This sounds like a problem for which Experiment Design software like "Design Expert" was built.