- #1
serbring
- 271
- 2
Hi all,
On a vehicle I recorded an output signal that is positive and it's variability is lead to the 30 input signals, but not all together at the same insant. Just by checking the pairwise correlation between signals in a time periond, I'm able to detect which input signals lead to the variability of the output signal. At first I tried with Pearson's correlation, but the correlation might be non linear and not always the two signals change monotically. For these reasons this correlation coefficient is not very helpful. Then, I tried with distance correlation and it works very well, but I miss the sign of the correlation that is really important to me. So what about using distance correlation parameter and the sign of Pearson's coefficition to detect the sign of the correlation (i.e. negative or positive correlation)? Thus I may be able to detect if an increase or a decrease of an input signal lead to an increase of the output signal. Any comment is appreciated.
Thanks
On a vehicle I recorded an output signal that is positive and it's variability is lead to the 30 input signals, but not all together at the same insant. Just by checking the pairwise correlation between signals in a time periond, I'm able to detect which input signals lead to the variability of the output signal. At first I tried with Pearson's correlation, but the correlation might be non linear and not always the two signals change monotically. For these reasons this correlation coefficient is not very helpful. Then, I tried with distance correlation and it works very well, but I miss the sign of the correlation that is really important to me. So what about using distance correlation parameter and the sign of Pearson's coefficition to detect the sign of the correlation (i.e. negative or positive correlation)? Thus I may be able to detect if an increase or a decrease of an input signal lead to an increase of the output signal. Any comment is appreciated.
Thanks