How can I import data from .xls using for Principal Component Analysis (PCA)?

10 次查看（过去 30 天）
Michelle
Michelle 2014-7-31
评论： Yaser Khojah 2019-4-19
Hello,
I have a worksheet in excel that contains the following data:
17 rows x 5 columns of data - the first row contains the title 3 of the variables (names of 3 different people), meanwhile column 1 contains the title for the 16 different possibilities. the internal matrix, which ends up being a 16x4, contains the statistic. I import the data into Matlab, but don't know if i should import as a matrix, cell array, or dataset. From there, how do I perform PCA? I want it to contain the names of the variables.
So far, I've performed this:
>> coeff = princomp(Cocktail1)
coeff =
0.0453 0.9976 0.0500 -0.0130
0.4413 -0.0662 0.8831 -0.1453
0.5300 -0.0067 -0.1275 0.8383
0.7227 -0.0172 -0.4489 -0.5253
However, the problem is it does not show me titles of the variables.
Thank you.

请先登录，再回答此问题。

采纳的回答

Shashank Prasanna
Michelle, your approach is correct. princomp (or pca which is a newer and recommended function to use) accepts only matrices as inputs, and returns the principal component coefficients as the first output.
It is important to note however, that pca transforms your data to a new coordinate space. This means your original variable names don't make much sense for your transformed data. The second output of pca is the transformed data which is a linear combination of your original data:
http://www.mathworks.com/help/stats/pca.html
If you want to simple assign some names to them, I recommend using Table as follows:
T = array2table(coeff,'VariableNames',{'Name1' 'Name2' 'Name3' 'Name4'})
  2 个评论
Michelle
Michelle 2014-8-1
Thank you for your answer! I understand that pca transforms the data onto a new coordinate space. But now, I want to plot the original data onto this new coordinate space which is where I am having trouble. Do you know how to plot the original data onto the biplot?
Yaser Khojah
Yaser Khojah 2019-4-19
Dear Shashank Prasanna I need help with PCA.
I'm looking for away to related PCA to the original data to find which variables have the highest impact. I understand there are
1) COEFF: Each column contains coefficients for one PC and they are in descending order
2) Score: Each column correspond to a PC and rows are related to observations
3) latent: the PC variances. I assume the first one is related to the first PC and so on
I have spent a lot of time trying to find a way to link the PC to the original data's variable. Is there anyway from looking at the PC and finds its related variables from the original data?

请先登录，再进行评论。

更多回答（0 个）

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

产品

Community Treasure Hunt

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

Start Hunting!

Translated by