how can I do PCA on a table ?

I have a 3000 x 18 table containing 18 variables (one per column) of both numerical and categorical type (6 categorical).

I would liket to do PCA on the whole dataset

but the command

[coeff,score]=pca(mytable) does not work.

any idea on how to proceed ?

Answer 1

[Star Strider]

The pca function only accepts single-precision or double-precision numerical arrays. You most likely need to use the table2array (link) function first, to get the numeric data out of your table and into a form that pca can use.

[LO]

Thanks,

I have tried and I got this error message

Error using pca (line 168)

The value of X must not be an object.

the table is now in string format. this is not suitable for the PCA function.

I tried to convert the matrix to numerical values using str2num but I got the message "Input must be a character vector or string scalar"

[Star Strider]

Unless you can extract just the numerical values, you cannot use the pca function.

[LO]

Got it, thanks. I thought categorical PCA was also possible with MATLAB.

Perhaps the answer is no. I will try to then exclude the non numerical variables and see if it works !

[Star Strider]

It should work in that instance.