can we implement PCA for lets 10 subjects 3 repeated trials with each trial 101 samples and variable vertical-GRF i.e. 101*30 input matrix

Principal Component Analysis

Answer 1

[Kartik Mittal]

My understanding of the problem is that you have 101 data points, each with 30 attributes? If that is the case, you can use PCA. But I am not sure you have a correct feature vector for what you are calling the data points (samples). If you have 10 subjects, with 3 trials, how are they attributive of one sample? In your case PCA would make sense if there is some correlation between trials, like stages of an experiment or something.

[imran mahmood]

Hi Kartik, thanks for your reply. My data set is 30trials (3 trials from each of the 10 subjects) and each trial has 101 samples. All trials were conducted under similar experimental conditions. My questions are: 1) Can I apply PCA with this data (rows:101, col:30) with one experimental condition. 2) What should be my input matrix (X); is my [X] = [(101 x 30)] to implement in MATLAB >>[coeff,score,latent,tsquared,explained,mu] = pca(X);

[Kartik Mittal]

1. It would make more sense if you could describe what your predictors are. As I stated in my answer above, "I am not sure you have a correct feature vector for what you are calling the data points (samples)." The interpretation : 3 trials x 10 subjects = 30 features, might not be correct when you are trying to use PCA. I don't understand fully what your workflow is.
2. You can use PCA if you are clear about 1.

[imran mahmood]

Hi Kartik thanks again, I am attaching the data set for my further understanding. In this data ankle-joint angle is measured as time-series waveforms. Three trials/waveforms were measured from each subject. There were total 10 subjects. I want to minimise the variance between the waveforms over the time span (101 samples). Can I use PCA?

[Kartik Mittal]

Thanks for the information, it makes the case clear that you will have repeated measures as your data points (given all the trials are with one experimental condition). Hence, PCA would not be an obvious choice for what you wish to do. Check this link - https://stats.stackexchange.com/questions/18617/can-i-do-a-pca-on-repeated-measures-for-data-reduction