regression method (in matlab)

Hi

I have two curves , one is measured one is calculated from cylinder. They are all angle dependent.

Angle->Pressure

Would you offer any matlab function that I calculate the regression with these two datas.

I am searching the qaulity of simulation regarding to measurement? The matrixes are like that

The Matrix A is (measured)

Angle     Pressure
-----     ------
1  3,35000000000000
2  3,33000000000000
3  3,31000000000000
4  3,35000000000000
5  3,31000000000000
....

the Matrix B (simulated)

Angle     Pressure
-----      --------
1  3,12302310000000
2  3,10574540000000
3  3,08854990000000
4  3,07221170000000
5  3,05751980000000

Best Regards.

[Star Strider]

What regression model are you considering?

[Orkun OZENER]

Dear Star Rider.,

I tried lineer regression model for obtaining how two data is connected to each other... The mean average values of two data sets are 3,256 and 3,4... But lineer regression gives 0,75 regression...So I I am really confused.. ı am open two all methods.

Answer 1

[Image Analyst]

Have you looked at polyfit()?

[Orkun OZENER]

Actually I am lookig a thing like the how the two series is connected to each other...I dont want to fit a polynom...

[Image Analyst]

You mean like a correlation coefficient, or RMS error, or PSNR? Usually you don't regress simulated vs. measured (i.e. predicted vs. observed). You could use scatter() to see them plotted against each other but usually you don't do a regression to fit a line through the points. Often/Usually you'll use PSNR to see how well your model/simulation did in predicting the observed values, at least in compressing and reconstructing data. Other fields may have other methods that are preferred, such as RMS error or average absolute difference or MSI or SSIM ( http://en.wikipedia.org/wiki/Structural_similarity)

Answer 2

[Star Strider]

You list one set of data as being ‘measured’ and the other as ‘simulated’. Are you comparing the fit of the simulation to the measured data? If so, then the best measure to describe their relationship might be the mean-square error:

MSE = mean((measured-simulated).^2);

The RMS error Image Analyst refers to is:

RMS = sqrt(MSE);

Answer 3

[Kazim Hanbay]

soru hakkında şunu söyleyebilirim;Her iki dağılımın olasılık yoğunluk fonksiyonlarını hesaplayıp daha sonra lineer reg. deneyebilirsiniz sanırım.Ayrıca size bir şey sormak istiyorum.Bir görüntünün korelasyon matrisini nasıl hesaplayabilirim?Daha sonra bu matrisin öz değer ve öz vektör analizini yapmam gerekiyor.

[Walter Roberson]

Mechanical translation of above (Google):

I can say about the question, the probability density functions of both distributions and then calculate the linear reg. I guess you can try. Also I want to ask you something. How do I calculate the correlation matrix of an image? Then I need to do an analysis of this matrix eigenvalues and eigenvectors.

[Orkun OZENER]

Dear Hanbay,Thanks for your answer. On the other hand I really dont know how you can do the thing that you asked.? Sorry . Best Regards.