Problems running PCA on CSV file

[andreaswallberg]

Dear @Zilong-Li

Thanks for a very interesting tool!

I have compiled the program and it seems to work. However, I have an issue with running it on a zstd-compressed CSV file. I get the following error:

[13/04/2025-20:58:41] program started 
[13/04/2025-20:58:41] start parsing CSV format compressed by ZSTD 
[13/04/2025-20:58:41] shape of input matrix (features x samples) is 12 x 572359
terminate called after throwing an instance of 'std::invalid_argument'
  what():  stod
Aborted (core dumped)

I am not sure the problem is with the compiled code or the input data.

I seem to be unable to find the example CSV file. Can you specify what a CSV file should look like (e..g. should we have a header line or not, are negative values tolerated, etc.) and share the one you use in your example.

Thanks!

[Zilong-Li]

Hi @andreaswallberg

Thanks for the info. I think the error is due to unexpected format. PCAone has limited support for CSV format, which has to inlcude no-header, no-quotes and only numeric entries.

If you download the example data package, you will find an example CSV file

wget http://popgen.dk/zilong/datahub/pca/example.tar.gz

Let me know if you have more bugs to report.

Best,
Zilong.

[andreaswallberg]

Many thanks!

Can we have floats and negative numbers, or only 0 and positive integers?

[Zilong-Li]

yes, you can! btw, I notice that your matrix is very thin, i.e. 12 x 572359. You will need the full SVD functionality by running --svd 3