MULTIVARIABLE REGRESSION

[naveenan88]

I have a come across a problem involving 6 different variables and 2 outputs. How do i go about finding the coefficients for this data. Does anyone know how i can tackle this problem? I thank you in advance.

 

[GregLocock]

What form of equation are you expecting? That is, a simple linear one, or what?

It would probably speed things no end if you were to post an example



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[IDS]

The on-line help on the Linest() function covers the procedure for a single unknown, or multiple unknowns.

If it doesn't cover what you want to do could you give us more specifics.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[katmar]

Excel has very powerful multivariable regression capabilities. Look in the Help file under "regression". It is part of the Data Analysis Pack that is not installed by default, so you may have to jump through a few hoops to get it to work.

Katmar Software
Engineering & Risk Analysis Software

http://katmarsoftware.com

 

[naveenan88]

GregLocock : Hi there. yeah i am new here. i am expecting a polynomial equation for this data. i have attached the data file along with this reply.i hope the data will give you an idea of my problem.thanks for the reply.

IDS:thank you for the info. i will look into it.

katmar: thanks for your suggestion.



Appreciate all for your replies.

Regards.

 

[magoo2]

If you look at your data, the AFR term is constant. I would expect that it won't be significant in the final result. So you only really have 5 factors.

 

[katmar]

I thought I would try to process your data, but it is not clear which are the dependent and independent variables. What form of equation are you looking to create?

Katmar Software
Engineering & Risk Analysis Software

http://katmarsoftware.com

 

[zelgar]

By turning on Autofilter, it appears that at least 5 the following components are the inputs: IVO, EVC, N, MAF and AFR.

Looking at IVO=14, EVC=45, there are several examples when N, MAF and AFR are the same (e.g., there are 2 records with N=3000, MAF=300, AFR=14.3) but slightly different anwers for NOx, SDNMEP and TORQUE.

Just from simple examination of the data, the value of N (3000 vs 1000) seams to affect NOx (N decreases, NOx Decreases) and MAF affects TORQUE (MAF decrease, TORQUE decreases)

 

[Clyde38]

See below;

http://www.j-walk.com/ss/excel/tips/tip101.htm

thread770-184726: Excel Curve Fit Coefficients
thread770-214997





[link

http://www.linkedin.com/in/clydehancock

]

[/url]

 

[magoo2]

I'm assuming the last 2 columns contain the outputs.

I can get a pretty good linear regression model for torque as a function of 4 of the inputs. R-squared is 92.5% so I don't see a need to go to a polynomial.

For the SDNMEP, the linear model is based on the same 4 inputs, but the R-squared is 46%. May want to look at a quadratic or cube model here.

 

[GregLocock]

IVO EVC N MAF AFR

NOx SDNMEP TORQUE

Isn't NOx an output as well?




Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[GregLocock]

Incidentally, using the excellent Antaeus I think it might be a little optimistic to rely on a linear regression.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[IDS]

Greg - I hadn't heard of Antaeus before (and I note that Wikipedia hasn't heard of it either).

Now being downloaded (free for academic and non-commercial use it says).

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[magoo2]

Greg,

The Antaeus output looks impressive. I'm assuming that it's an interaction matrix. Will it also permit you to develop a regression equation?

Like I said, a linear model works fine if torque is an output.

For the SDNMEP, it's more complicated. I didn't bother to pursue it since I only assumed it was an output. It seems more logical to me to start with the inputs and finish with the outputs.

Maybe the newbee will clarify what are the inputs and what are the outputs. Or maybe this is homework!

 

[naveenan88]

magoo: thanks for ur info. it gave me some idea

katmar: the IVO, EVC,N,MAF,AFR and NOx are my independent variables. the other 2 are the dependent variables. the task is that i'll need to find out what form of equation is suitable for this data.

greg: y do you think say that NOx is an output?

THE PROBLEM I AM FACING IS THAT I DO NOT KNOW HOW TO CONVERT ALL THE INPUTS INTO A VECTOR ,SO THAT I COULD PLOT IT IN A GRAPH.CAN ANYONE SHOW ME HOW TO GO ABOUT IT?

I THANK U IN ADVANCE

 

[magoo2]

Sure sounds like homework to me. Sorry.

 

[GregLocock]

Oh go on, do tell me how you were injecting NOx into the engine.

Never the less, you've been given several sufficient links for a regression in Excel, and some comments indicating that a linear model may not suffice.

Short of doing your homework for you, I think you need to ask specific questions having actually tried to follow the advice thus far.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?