search

precimed / mixer

Causal Mixture Model for GWAS summary statistics
GNU General Public License v3.0
59 stars 17 forks source link

The link to download .ld and .snp files no longer exists #63

Closed yyy-nkl closed 1 year ago

ofrei commented 1 year ago

@yyy-nkl the files are available here https://github.com/comorment/mixer/tree/main/reference/ldsc/1000G_EUR_Phase3_plink please let me know if some old references were not updated

yyy-nkl commented 1 year ago

Thank you for your help!  I have another question.  Why need to generate 20 repeats of file.rep${SLURM_ARRAY_TASK_ID}.snps files? Why is the number of repetitions not some other number? I found that it takes a long time to fit the model in univariate analysis and it takes 24h or more, so I would like to ask if reducing the number of repeats will have any impact? Thank you in advance for your help!

G.Y. @.***

 

------------------ Original ------------------ From:  "Oleksandr @.>; Send time: Thursday, Dec 29, 2022 6:22 PM @.>; @.>; @.>; Subject:  Re: [precimed/mixer] The link to download .ld and .snp files no longer exists (Issue #63)

@yyy-nkl the files are available here https://github.com/comorment/mixer/tree/main/reference/ldsc/1000G_EUR_Phase3_plink please let me know if some old references were not updated

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

ofrei commented 1 year ago

@yyy-nkl the results are averaged across 20 repeats to provide more stable point estimate overall (in case of convergence issues, e.g. if a single run lead to an outlier estimates), and get an estimate of standard deviation of those point estimates. There is nothing special about N=20 repeats, you may consider using e.g. 10 repeats, though I think it's too few for estimating a standard deviation (even 20 is quite low).

yyy-nkl commented 1 year ago

Thank you for your patient explanation!

G.Y. @.***

 

------------------ Original ------------------ From:  "Oleksandr @.>; Send time: Thursday, Dec 29, 2022 7:46 PM @.>; @.>; @.>; Subject:  Re: [precimed/mixer] The link to download .ld and .snp files no longer exists (Issue #63)

@yyy-nkl the results are averaged across 20 repeats to provide more stable point estimate overall (in case of convergence issues, e.g. if a single run lead to an outlier estimates), and get an estimate of standard deviation of those point estimates. There is nothing special about N=20 repeats, you may consider using e.g. 10 repeats, though I think it's too few for estimating a standard deviation (even 20 is quite low).

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>