search

leelabsg / SKAT

Sequence kernel association test (SKAT)
41 stars 16 forks source link

Memory requirements from creating unused diagonal matrix #4

Closed jmerkin closed 2 years ago

jmerkin commented 5 years ago

In the function SKAT_Optimal_Logistic_VarMatching, there's a call to create a diagonal matrix that is never used. This matrix is never used but the memory allocation can grow quite large.

Line 303 in the file SKAT_Optimal_VarMatching.R: D = diag(pi_1)

can be removed without seemingly affecting anything downstream.

ZhangchenZhao commented 4 years ago

Thanks for this great suggestion! We already remove this line. Please try our updated version, which can be found in "Sparse_Version" branch. Now we can support both matrix and sparse matrix.

complexgenome commented 4 years ago

@ZhangchenZhao Hi, sorry to hijack this thread. Can you direct to the "Sparse_Version"? I'd like to use it but unable to find as is the link provided in paper

Robust gene-based test, https://github.com/leeshawn/SKAT/tree/ Sparse_Version

UK Biobank Whole-Exome Sequence Binary Phenome Analysis with Robust Region-Based Rare-Variant Test https://www.sciencedirect.com/science/article/pii/S0002929719304331

Thanks.

ZhangchenZhao commented 4 years ago

Hi Sanjeev,

Thanks for your email. You can find updated SKAT through this link https://github.com/leeshawn/SKAT/tree/Sparse_Version. Later this month we will also upload this SKAT version to CRAN.

Let me know if you have any questions.

Thanks, Zhangchen

On Mon, Jan 13, 2020 at 11:19 AM Sanjeev notifications@github.com wrote:

@ZhangchenZhao Hi, sorry to hijack this thread. Can you direct to the "Sparse_Version"? I'd like to use it but unable to find as is the link provided in paper

Robust gene-based test, https://github.com/leeshawn/SKAT/tree/ Sparse_Version

UK Biobank Whole-Exome Sequence Binary Phenome Analysis with Robust Region-Based Rare-Variant Test https://www.sciencedirect.com/science/article/pii/S0002929719304331

Thanks.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.