DReichLab / DReichLab/EIG

A low number of iterations in fastmode decreases accuracy

Open
#5 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C
Stars
207
Forks
61
PR merge metrics
No merged PRs in 30d

Description

Hi, you may wish to increase the default number of iterations in fastpca (fastiter), for some datasets the default is too low, resulting in inaccurate estimation of eigenvectors (I checked this by comparing the non-fastmode results with the fastmode, and verified with several other SVD implementations).

For example with the default iterations:

> diag(cor(P1, P2))
        V2         V3         V4         V5         V6         V7         V8
 0.9985658  0.9909673  0.9819172 -0.9131868  0.9447627  0.9555776 -0.8435918
        V9        V10        V11
-0.7538453 -0.6555316  0.8961565

With fastiter: 100

> diag(cor(P1, P2))
 V2  V3  V4  V5  V6  V7  V8  V9 V10 V11
  1   1   1  -1  -1  -1  -1   1   1   1

where P1 and P2 are the top ten eigenvectors from non-fastmode and fastmode, respectively. The dataset consists of 5300 individuals with 3700 LD-thinned SNPs.

I'm using v6.1.1.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by locating the fastpca fastmode implementation and the fastiter default. Compare fastmode eigenvectors with non-fastmode results on the reported dataset, including fastiter 100; done when the default produces accurate results comparable to the non-fastmode and other SVD implementations.

Written by the indexing model from the issue text.

Assessment

Tech stack
c
Domain
data
Issue type
Bug
Difficulty
3/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.