Why is my coeff variable blank when I run PCA?

8 次查看（过去 30 天）
Kelly Thompson
Kelly Thompson 2019-5-11
回答： Paras Gupta 2024-7-19
So I would like to conduct PCA on a dataset I have that is of sample size 23 and feature size 813 (nxp = 23x813). However, when I run the pca function, the coeff variable returned is blank. why is this the case? I am aware my sample size: feature size ratio is very low, but the function returned a 813x22 coefficient matrix when I ran it on my data before I processed it (transformed nonnormally distributed data & feature scaling) so I know it is not because of unsuitable data size.
I have attached a copy of the data I am trying to use if anyone is able to help; thanks!
[coeff, ~, latent] = pca(deltaEBC);
UPDATE: I was feature scaling by taking feature X, subtracting the mean of X, and dividing by the standard deviation of X. I changed this to simply subtracting the mean and not dividing by the std, and now the pca function does return a 813 x 22 coeff I was looking for. I just don't know why?

请先登录，再回答此问题。

回答（1 个）

Paras Gupta
Paras Gupta 2024-7-19
Hi Kelly,
I understand that you encountered an issue with the "pca" function in MATLAB where the "coeff" variable returned was blank.
This issue is due to the presence of NaN columns in your dataset, where all rows in certain columns contain NaN.
By default, the "pca" function in MATLAB performs the action specified by the 'Rows','complete' name-value pair argument, which removes rows with NaN values before calculation. Please refer the following documentation link for more information:
https://www.mathworks.com/help/stats/pca.html#bth9ibe-Rows
You may need to handle the NaN columns and remove them. You can refer the documentation on the "rmmissing" function - https://www.mathworks.com/help/matlab/ref/rmmissing.html
deltaEBC_new = rmmissing(deltaEBC, 2);
However, if you do do not want to remove missing entries from your data, you can use the Alternating least squares (ALS) algorithm for PCA in matlab which better handles missing values. You can refer the folllowing link to select this algorithm - https://www.mathworks.com/help/stats/pca.html#bth9ibe-Algorithm
[coeff, ~, latent] = pca(deltaEBC, Algorithm="als");
Hope this solves the issue.

类别

在 Help Center 和 File Exchange 中查找有关 Dimensionality Reduction and Feature Extraction 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by