MATLAB Answers

0

Calculate different correlation coefficients

Asked by Marc Gebhardt on 8 Jul 2019
Latest activity Answered by Sai Bhargav Avula on 30 Jul 2019
Hi guys,
i want to calculate correlation coefficients for some dataseries.
The first groupwise representation is shown below. All datapoints are clustered into groups, if they are fulfilling specific attributes. Datapoints without any group are in orange. The functions between y and x can be seen groupwise (y=x^2 for blue and yellow, and a small-regional circleplot for purple and green).
Bild2.png
The calculated correlation coefficients are -0.87, -0.89, -0.25 and -0.05. This is comprehensible regarding the data in the plot.
Therefore I was using the corrcoef function.
For another dataset, same procedure is done.
Bild1.png
There arent any functional relations (in my opinion), but the correlation coefficients are -0.86 0.78 for example.
I didnt get these high correlation regarding this plot. What function can I use to directly decide if one group of points is correlating or not?

  0 Comments

Sign in to comment.

1 Answer

Answer by Sai Bhargav Avula on 30 Jul 2019

In the second figure the data is non-linearly correlated and are of non-parametric distribution. In such cases rank correlation would work. The corrcoeff function analyses the linear relation may not derive the correlation as expected. I would suggest using Kendall's tau or Spearman's rho which are available in the function corr, which is the part of statistics toolbox.
Refer to the following documentation of the corr function to know the other options available

  0 Comments

Sign in to comment.