# Thread: Regression analysis with independant variable (error) and multiple dependant variable

1. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

A multiple linear regression model is linear in the parameters (the "betas") but not necessarily linear in the variables: So this is a linear model:

y = beta0 + beta1*speed + beta2*speed^2 + beta3*log(voltage) + beta4*sin(angle) + beta5*temp + epsilon

Where y is the dependent variable (Eloks positional deviation) and epsilon a random disturbance term (that in statistical jargon often is called an "error term").

In most statistical software you just "mention what explanatory variables you want to have". (In spss you click in then in a box for independent variables). What is Eloks software?

In R you do:

Code:
``lm( y ~ speed + speed^2 + log(voltage) + sin(angle) + temp)``
If Elok is completely unfamiliar with multiple regression I would sugest to consider an statistical experimental plan. It is better with a carefully design plan with 20 - 30 experimental points, than 1000 messy observational points. I suggest to vary two variables, like speed and voltage, each on high and low values.

2. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

I'm using Matlab but I can open to others software. I got a backgound in programming so I don't mind learning software of language.

But I'm not sure I follow GretaGarbo's example, but the link I posted in another post is giving me a lot of help so far.

3. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

Tell us about your awesome 3-D printer? What do you typically print?

4. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

I can't, sorry :P

5. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

Anyway, back to the subject, does anyone of you know how to make this sort of "Polynomial Multiple Regression analysis" like this one? Just show me how I can calculate the curved grid of the picture from a database of [10000 x 3] value. It's all I need.

6. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

It looks like a 3-D correlation or Spline. Have you looked up Spline Regression yet?

7. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

Originally Posted by hlsmith
It looks like a 3-D correlation or Spline. Have you looked up Spline Regression yet?
Will do thanks!

I also have a friend that told me about a tool on MatLab called Curve fitting tool. Will check that also.

8. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

oh, he just want to know how to make that graph of the linear regression with polynomials. Not how to analyse the data or do significance test or anything.

But matlab is a quite old program so I don't think anybody here is particularly enthusiastic about that one. Here is an R link.

If the results matter to you, (I agree with PeterFlom) go and talk to a statistician.

9. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

I've read a lot about R and it's performance seem abysimal even compared to MatLab. I was thinking about either Python (anaconda package seem quite awesome) or another program I don't remember the name that my university is offering for free (StatGraphic or something with stat in the name).

10. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

R - abysmal?

Such a simple computation can be done in much less than a second.

But you can download it for free, and start working with it faster than it takes you to walk to, or call, the university administrator to ask for allowance to use statgraphics. R has got 6000 extra packages. Do you think statgraphics has got 60? But you Elok only need the base version and and to study the basic of multiple linear regression.

11. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

I dont know GG that linked graph is not as cool as his attached image.

12. ## Re: Regression analysis with independant variable (error) and multiple dependant vari

But even if each of those individuals parameters doesn't explain my error perfectly, I highly confident a combination of 2-3-4 parameter would have a really high r2.
Don't forget the possible interactions between parameters.

Page 2 of 2 First 1 2

 Tweet