Hi, I have been trying to analyse a binary trait in UK biobank first

UK biobank analysis,about weizhouumich/saige

Comments (21)

weizhouUMICH commented on August 23, 2024

Hi Nicola,
This error message is usally due to the non-invertible covariate matrix.
Which version of SAIGE are you using? The newer versions > 0.23 would have solved this issue automatically.
Thanks,
Wei

from saige.

npirastu commented on August 23, 2024

Hi Wei, I am actually using version 0.26. N Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG [cid:[email protected]][cid:[email protected]] Il giorno 15 feb 2018, alle ore 15:52, weizhouUMICH <[email protected]<mailto:[email protected]>> ha scritto: Hi Nicola, This error message is usally due to the non-invertible covariate matrix. Which version of SAIGE are you using? The newer versions > 0.23 would have solved this issue automatically. Thanks, Wei — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XClIjkd_8rYAJLwfPb54HiO8zw-dGks5tVFKrgaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

weizhouUMICH commented on August 23, 2024

Interesting. Could you please have a check to see whether the second non-genetic covariate in your formula has identical values for the tested samples?

from saige.

npirastu commented on August 23, 2024

I am sure that’s not the case. I have tryed even y~1 and still get the same problem. Unfortunately I can code only in R so I can’t debug it myself.

…

________________________________ From: weizhouUMICH <[email protected]> Sent: 15 February 2018 16:54:19 To: weizhouUMICH/SAIGE Cc: PIRASTU Nicola; Author Subject: Re: [weizhouUMICH/SAIGE] UK biobank analysis (#17) Interesting. Could you please have a check to see whether the second non-genetic covariate in your formula has identical values for the tested samples? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XCsbJHRJen87MmWMPWMa--_UiWfJxks5tVGE7gaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

npirastu commented on August 23, 2024

Dear Wei, I have looked at the error and it this is what I am getting. [1] "Start reading genotype plink file here" nbyte: 25 nbyte: 25 reserve: 288117 M: 10671, N: 99 size of genoVecofPointers: 1 here time: 44.2 [1] "Genotype reading is done" inital tau is 1 0.5 iGet_Coef: 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 error: inv_sympd(): matrix is singular or not positive definite Error in getCoefficients(Y, X, W, tau, maxiter = maxiterPCG, tol = tolPCG) : inv_sympd(): matrix is singular or not positive definite Timing stopped at: 0.268 0.052 0.47 I am not sure if it helps understanding what may be going on. Many thanks Nicola Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG [cid:[email protected]][cid:[email protected]] Il giorno 15 feb 2018, alle ore 16:54, weizhouUMICH <[email protected]<mailto:[email protected]>> ha scritto: Interesting. Could you please have a check to see whether the second non-genetic covariate in your formula has identical values for the tested samples? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XCsbJHRJen87MmWMPWMa--_UiWfJxks5tVGE7gaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

weizhouUMICH commented on August 23, 2024

Dear Nicola,

Thank you so much for providing more details! Sorry for dropping the ball for a while. I'm happy to help.
You mentioned that when no covariates were used, you saw the same error message.
May I ask how many covariates are used for the jobs shown in the log file?

Thanks,
Wei

from saige.

npirastu commented on August 23, 2024

For this model I used one covariate which is not monomorphic if you are wondering. Dear Wei, I think the issue is that kinship matrices are often not exactly positive definite and the inv_sympd() does not work. I have tried to change it to inv() and recompile the package but it appears that one of the third party cget package some files are not there and there are links in it’s place which makes the compiling crush. I have run this dataset before with GCTA which worked, of course the probit model is horrible so I was very happy to be able to run a proper logit model. Many thanks Nicola Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG [cid:[email protected]][cid:[email protected]] Il giorno 07 mar 2018, alle ore 04:57, weizhouUMICH <[email protected]<mailto:[email protected]>> ha scritto: Dear Nicola, Thank you so much for providing more details! Sorry for dropping the ball for a while. I'm happy to help. You mentioned that when no covariates were used, you saw the same error message. May I ask how many covariates are used for the jobs shown in the log file? Thanks, Wei — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XChbNWmvjszW8WQ-EdjKLO0IZwqmJks5tb2jKgaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

weizhouUMICH commented on August 23, 2024

Dear Nicola, Thank you for letting me know! I have made a version 0.26.1, in which the inv_sympd() is replaced by inv(). https://www.dropbox.com/sh/pcql2hrdid63czc/AAC7Lm4wkbtqK9zththZWCGFa?dl=0 Could you please try it out? Btw, if you don't mind sharing the error message for the compiling, I would take a look to see if I can fix it. Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor

…

On Wed, Mar 7, 2018 at 5:38 AM, npirastu ***@***.***> wrote: For this model I used one covariate which is not monomorphic if you are wondering. Dear Wei, I think the issue is that kinship matrices are often not exactly positive definite and the inv_sympd() does not work. I have tried to change it to inv() and recompile the package but it appears that one of the third party cget package some files are not there and there are links in it’s place which makes the compiling crush. I have run this dataset before with GCTA which worked, of course the probit model is horrible so I was very happy to be able to run a proper logit model. Many thanks Nicola Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG ***@***.***[cid: ***@***.*** Il giorno 07 mar 2018, alle ore 04:57, weizhouUMICH < ***@***.******@***.***>> ha scritto: Dear Nicola, Thank you so much for providing more details! Sorry for dropping the ball for a while. I'm happy to help. You mentioned that when no covariates were used, you saw the same error message. May I ask how many covariates are used for the jobs shown in the log file? Thanks, Wei — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://github.com/ weizhouUMICH/SAIGE#17#issuecomment-371022949>, or mute the thread< https://github.com/notifications/unsubscribe-auth/Ai0XChbNWmvjszW8WQ- EdjKLO0IZwqmJks5tb2jKgaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJfdL7YX_U_dpACd26IssQiZotRcI3Wnks5tb7ibgaJpZM4SHBvO> .

from saige.

npirastu commented on August 23, 2024

Dear Wei, sorry for taking so long to respond. The problem with the inversion of the matrix seems to be fixed. I am now getting: CVD is a binary trait Call: glm(formula = formula.new, family = binomial, data = data.new) Coefficients: minus1 Coronary.heart.disease_8 3.0067 0.2673 Degrees of Freedom: 134938 Total (i.e. Null); 134936 Residual Null Deviance: 187100 Residual Deviance: 52000 AIC: 52010 [1] "Start reading genotype plink file here" nbyte: 38183 nbyte: 33735 reserve: 360007520 M: 10671, N: 152732 size of genoVecofPointers: 1 here time: 47915 [1] "Genotype reading is done" inital tau is 1 0.5 iGet_Coef: 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 Tau: [1] 1.0 0.5 Fixed-effect coefficients: [,1] [1,] NaN [2,] NaN Error in if (max(abs(alpha - alpha0)/(abs(alpha) + abs(alpha0) + tol.coef)) < : missing value where TRUE/FALSE needed Timing stopped at: 73.18 9.298 52.25 I am not sure what this means. Thanks a lot for all the help. Nicola Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG Il giorno 08 mar 2018, alle ore 16:40, weizhouUMICH <[email protected]<mailto:[email protected]>> ha scritto: Dear Nicola, Thank you for letting me know! I have made a version 0.26.1, in which the inv_sympd() is replaced by inv(). https://www.dropbox.com/sh/pcql2hrdid63czc/AAC7Lm4wkbtqK9zththZWCGFa?dl=0 Could you please try it out? Btw, if you don't mind sharing the error message for the compiling, I would take a look to see if I can fix it. Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor

On Wed, Mar 7, 2018 at 5:38 AM, npirastu ***@***.******@***.***>> wrote: For this model I used one covariate which is not monomorphic if you are wondering. Dear Wei, I think the issue is that kinship matrices are often not exactly positive definite and the inv_sympd() does not work. I have tried to change it to inv() and recompile the package but it appears that one of the third party cget package some files are not there and there are links in it’s place which makes the compiling crush. I have run this dataset before with GCTA which worked, of course the probit model is horrible so I was very happy to be able to run a proper logit model. Many thanks Nicola Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG ***@***.***[cid: ***@***.******@***.***>] Il giorno 07 mar 2018, alle ore 04:57, weizhouUMICH < ***@***.******@***.******@***.***>> ha scritto: Dear Nicola, Thank you so much for providing more details! Sorry for dropping the ball for a while. I'm happy to help. You mentioned that when no covariates were used, you saw the same error message. May I ask how many covariates are used for the jobs shown in the log file? Thanks, Wei — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://github.com/ weizhouUMICH/SAIGE#17#issuecomment-371022949>, or mute the thread< https://github.com/notifications/unsubscribe-auth/Ai0XChbNWmvjszW8WQ- EdjKLO0IZwqmJks5tb2jKgaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJfdL7YX_U_dpACd26IssQiZotRcI3Wnks5tb7ibgaJpZM4SHBvO> .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XCtidP15i-s9VD4IZCOEk39FMxM1Rks5tcV8agaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

weizhouUMICH commented on August 23, 2024

Hi Nicola, Thank you for letting me know! It seems the inv() still does not work correctly. I have added some command to output the "cov", "Sigma_iX", and "X" after applying the inv() function. You can find the version 0.26.6 in the folder testingVersion/forNicola https://www.dropbox.com/sh/pcql2hrdid63czc/AAC7Lm4wkbtqK9zththZWCGFa?dl=0 Could you please run your job using this script and let me know how these three variables look? Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor

…

On Fri, Mar 16, 2018 at 8:13 AM, npirastu ***@***.***> wrote: Dear Wei, sorry for taking so long to respond. The problem with the inversion of the matrix seems to be fixed. I am now getting: CVD is a binary trait Call: glm(formula = formula.new, family = binomial, data = data.new) Coefficients: minus1 Coronary.heart.disease_8 3.0067 0.2673 Degrees of Freedom: 134938 Total (i.e. Null); 134936 Residual Null Deviance: 187100 Residual Deviance: 52000 AIC: 52010 [1] "Start reading genotype plink file here" nbyte: 38183 nbyte: 33735 reserve: 360007520 M: 10671, N: 152732 size of genoVecofPointers: 1 here time: 47915 [1] "Genotype reading is done" inital tau is 1 0.5 iGet_Coef: 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 Tau: [1] 1.0 0.5 Fixed-effect coefficients: [,1] [1,] NaN [2,] NaN Error in if (max(abs(alpha - alpha0)/(abs(alpha) + abs(alpha0) + tol.coef)) < : missing value where TRUE/FALSE needed Timing stopped at: 73.18 9.298 52.25 I am not sure what this means. Thanks a lot for all the help. Nicola Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG Il giorno 08 mar 2018, alle ore 16:40, weizhouUMICH < ***@***.******@***.***>> ha scritto: Dear Nicola, Thank you for letting me know! I have made a version 0.26.1, in which the inv_sympd() is replaced by inv(). https://www.dropbox.com/sh/pcql2hrdid63czc/AAC7Lm4wkbtqK9zththZWCGFa?dl=0 Could you please try it out? Btw, if you don't mind sharing the error message for the compiling, I would take a look to see if I can fix it. Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor On Wed, Mar 7, 2018 at 5:38 AM, npirastu ***@***.***<mailto: ***@***.***>> wrote: > For this model I used one covariate which is not monomorphic if you are > wondering. > > Dear Wei, > > I think the issue is that kinship matrices are often not exactly positive > definite and the inv_sympd() does not work. > I have tried to change it to inv() and recompile the package but it > appears that one of the third party cget package some files are not there > and there are links in it’s place which makes the compiling crush. > I have run this dataset before with GCTA which worked, of course the > probit model is horrible so I was very happy to be able to run a proper > logit model. > > Many thanks > > Nicola > > > > > > > > Nicola Pirastu > Chancellor’s Fellow. > Usher Institute > The University of Edinburgh > Room 3.732, Doorway 1 > Old Medical School > Teviot Place > Edinburgh EH8 9AG > > > ***@***.***[cid: > ***@***.***<mailto:F ***@***.***>] > > Il giorno 07 mar 2018, alle ore 04:57, weizhouUMICH < > ***@***.******@***.***><mailto: ***@***.***>> ha scritto: > > > Dear Nicola, > > Thank you so much for providing more details! Sorry for dropping the ball > for a while. I'm happy to help. > You mentioned that when no covariates were used, you saw the same error > message. > May I ask how many covariates are used for the jobs shown in the log file? > > Thanks, > Wei > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub<https://github.com/ > weizhouUMICH/SAIGE#17#issuecomment-371022949>, or mute the thread< > https://github.com/notifications/unsubscribe-auth/Ai0XChbNWmvjszW8WQ- > EdjKLO0IZwqmJks5tb2jKgaJpZM4SHBvO>. > > > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#17 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AJfdL7YX_U_ dpACd26IssQiZotRcI3Wnks5tb7ibgaJpZM4SHBvO> > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://github.com/ weizhouUMICH/SAIGE#17#issuecomment-371545068>, or mute the thread< https://github.com/notifications/unsubscribe-auth/Ai0XCtidP15i- s9VD4IZCOEk39FMxM1Rks5tcV8agaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJfdL2jGQuVAvei9BCs3JIV7XZtRRdXrks5te6x2gaJpZM4SHBvO> .

from saige.

npirastu commented on August 23, 2024

Hi Wei, here’s the output. I understand now why it would not work, but not sure why this happens. re.coef$cov: NaN NaN 0 NaN re.coef$Sigma_iX: NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN X: -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -0.07974977 -0.788817 -0.4151616 1.38986 0.01953004 -0.0393048 -0.2328933 -3.209997 0.6207993 0.8942355 -0.3338062 -1.873469 0.1249467 1.33211 0.5184882 1.669455 -0.1534982 1.057121 -1.696128 0.8965702 -0.6077298 -0.09570841 -1.096816 -1.559142 -0.06841314 0.1484303 0.5139157 -0.4743558 -0.17983 -1.139673 0.7231793 0.2152666 -1.029023 -0.2215924 -0.2958078 -0.6478647 -0.4384482 0.07254195 0.6202964 -0.4080381 2.317994 1.11522 -1.025295 -0.108153 0.007951344 0.4605968 -0.7771759 1.254086 0.4339252 1.79365 -0.144825 -1.043396 -0.8638546 -0.3958679 0.5721337 1.182113 0.3987589 0.7440897 -1.116444 -1.595013 -0.09620822 0.3846312 0.306187 0.593701 -0.8289956 1.81365 0.1594957 -0.3564891 -0.1736728 -1.733334 0.08284035 0.4424332 -1.968456 0.08998163 1.071192 -1.839723 0.0383429 1.171515 -0.7542824 -0.003146338 -1.141178 1.877035 1.096546 2.321842 -0.326046 -0.6389182 0.8493898 0.5236789 1.21312 0.1767615 -0.8480948 0.6114235 -0.6352869 -1.958003 1.269004 0.4907394 0.172751 -0.8599504 0.4635506 Best N Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG [cid:[email protected]][cid:[email protected]] Il giorno 16 mar 2018, alle ore 15:42, weizhouUMICH <[email protected]<mailto:[email protected]>> ha scritto: Hi Nicola, Thank you for letting me know! It seems the inv() still does not work correctly. I have added some command to output the "cov", "Sigma_iX", and "X" after applying the inv() function. You can find the version 0.26.6 in the folder testingVersion/forNicola https://www.dropbox.com/sh/pcql2hrdid63czc/AAC7Lm4wkbtqK9zththZWCGFa?dl=0 Could you please run your job using this script and let me know how these three variables look? Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor

On Fri, Mar 16, 2018 at 8:13 AM, npirastu ***@***.******@***.***>> wrote: Dear Wei, sorry for taking so long to respond. The problem with the inversion of the matrix seems to be fixed. I am now getting: CVD is a binary trait Call: glm(formula = formula.new, family = binomial, data = data.new) Coefficients: minus1 Coronary.heart.disease_8 3.0067 0.2673 Degrees of Freedom: 134938 Total (i.e. Null); 134936 Residual Null Deviance: 187100 Residual Deviance: 52000 AIC: 52010 [1] "Start reading genotype plink file here" nbyte: 38183 nbyte: 33735 reserve: 360007520 M: 10671, N: 152732 size of genoVecofPointers: 1 here time: 47915 [1] "Genotype reading is done" inital tau is 1 0.5 iGet_Coef: 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 Tau: [1] 1.0 0.5 Fixed-effect coefficients: [,1] [1,] NaN [2,] NaN Error in if (max(abs(alpha - alpha0)/(abs(alpha) + abs(alpha0) + tol.coef)) < : missing value where TRUE/FALSE needed Timing stopped at: 73.18 9.298 52.25 I am not sure what this means. Thanks a lot for all the help. Nicola Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG Il giorno 08 mar 2018, alle ore 16:40, weizhouUMICH < ***@***.******@***.******@***.***>> ha scritto: Dear Nicola, Thank you for letting me know! I have made a version 0.26.1, in which the inv_sympd() is replaced by inv(). https://www.dropbox.com/sh/pcql2hrdid63czc/AAC7Lm4wkbtqK9zththZWCGFa?dl=0 Could you please try it out? Btw, if you don't mind sharing the error message for the compiling, I would take a look to see if I can fix it. Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor On Wed, Mar 7, 2018 at 5:38 AM, npirastu ***@***.******@***.***><mailto: ***@***.******@***.***>>> wrote: > For this model I used one covariate which is not monomorphic if you are > wondering. > > Dear Wei, > > I think the issue is that kinship matrices are often not exactly positive > definite and the inv_sympd() does not work. > I have tried to change it to inv() and recompile the package but it > appears that one of the third party cget package some files are not there > and there are links in it’s place which makes the compiling crush. > I have run this dataset before with GCTA which worked, of course the > probit model is horrible so I was very happy to be able to run a proper > logit model. > > Many thanks > > Nicola > > > > > > > > Nicola Pirastu > Chancellor’s Fellow. > Usher Institute > The University of Edinburgh > Room 3.732, Doorway 1 > Old Medical School > Teviot Place > Edinburgh EH8 9AG > > > ***@***.***[cid: > ***@***.******@***.***><mailto:F ***@***.******@***.***>>] > > Il giorno 07 mar 2018, alle ore 04:57, weizhouUMICH < > ***@***.******@***.******@***.***><mailto: ***@***.******@***.***>>> ha scritto: > > > Dear Nicola, > > Thank you so much for providing more details! Sorry for dropping the ball > for a while. I'm happy to help. > You mentioned that when no covariates were used, you saw the same error > message. > May I ask how many covariates are used for the jobs shown in the log file? > > Thanks, > Wei > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub<https://github.com/ > weizhouUMICH/SAIGE#17#issuecomment-371022949>, or mute the thread< > https://github.com/notifications/unsubscribe-auth/Ai0XChbNWmvjszW8WQ- > EdjKLO0IZwqmJks5tb2jKgaJpZM4SHBvO>. > > > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#17 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AJfdL7YX_U_ dpACd26IssQiZotRcI3Wnks5tb7ibgaJpZM4SHBvO> > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://github.com/ weizhouUMICH/SAIGE#17#issuecomment-371545068>, or mute the thread< https://github.com/notifications/unsubscribe-auth/Ai0XCtidP15i- s9VD4IZCOEk39FMxM1Rks5tcV8agaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJfdL2jGQuVAvei9BCs3JIV7XZtRRdXrks5te6x2gaJpZM4SHBvO> .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XCtAFNB9qvqB13e-RaspPQbzyLY6_ks5te91MgaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

weizhouUMICH commented on August 23, 2024

Hi Nicola, Thank you so much for your quick feedback! It is likely that there is the perfect separation in the model. (one of the covariate can perfectly separate 0 or 1 for the phenotype) To test this, I made the program output the "mu", which will close to 0 or 1 if there is perfect separation. Could you please have a test run again? The program version is still the 0.26.6 in the same folder. It has been just updated Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor

…

On Fri, Mar 16, 2018 at 11:53 AM, npirastu ***@***.***> wrote: Hi Wei, here’s the output. I understand now why it would not work, but not sure why this happens. re.coef$cov: NaN NaN 0 NaN re.coef$Sigma_iX: NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN X: -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -0.07974977 -0.788817 -0.4151616 1.38986 0.01953004 -0.0393048 -0.2328933 -3.209997 0.6207993 0.8942355 -0.3338062 -1.873469 0.1249467 1.33211 0.5184882 1.669455 -0.1534982 1.057121 -1.696128 0.8965702 -0.6077298 -0.09570841 -1.096816 -1.559142 -0.06841314 0.1484303 0.5139157 -0.4743558 -0.17983 -1.139673 0.7231793 0.2152666 -1.029023 -0.2215924 -0.2958078 -0.6478647 -0.4384482 0.07254195 0.6202964 -0.4080381 2.317994 1.11522 -1.025295 -0.108153 0.007951344 0.4605968 -0.7771759 1.254086 0.4339252 1.79365 -0.144825 -1.043396 -0.8638546 -0.3958679 0.5721337 1.182113 0.3987589 0.7440897 -1.116444 -1.595013 -0.09620822 0.3846312 0.306187 0.593701 -0.8289956 1.81365 0.1594957 -0.3564891 -0.1736728 -1.733334 0.08284035 0.4424332 -1.968456 0.08998163 1.071192 -1.839723 0.0383429 1.171515 -0.7542824 -0.003146338 -1.141178 1.877035 1.096546 2.321842 -0.326046 -0.6389182 0.8493898 0.5236789 1.21312 0.1767615 -0.8480948 0.6114235 -0.6352869 -1.958003 1.269004 0.4907394 0.172751 -0.8599504 0.4635506 Best N Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG ***@***.***[cid: ***@***.*** Il giorno 16 mar 2018, alle ore 15:42, weizhouUMICH < ***@***.******@***.***>> ha scritto: Hi Nicola, Thank you for letting me know! It seems the inv() still does not work correctly. I have added some command to output the "cov", "Sigma_iX", and "X" after applying the inv() function. You can find the version 0.26.6 in the folder testingVersion/forNicola https://www.dropbox.com/sh/pcql2hrdid63czc/AAC7Lm4wkbtqK9zththZWCGFa?dl=0 Could you please run your job using this script and let me know how these three variables look? Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor On Fri, Mar 16, 2018 at 8:13 AM, npirastu ***@***.***< ***@***.***>> wrote: > Dear Wei, > > sorry for taking so long to respond. > > The problem with the inversion of the matrix seems to be fixed. > I am now getting: > > > CVD is a binary trait > > Call: glm(formula = formula.new, family = binomial, data = data.new) > > Coefficients: > minus1 Coronary.heart.disease_8 > 3.0067 0.2673 > > Degrees of Freedom: 134938 Total (i.e. Null); 134936 Residual > Null Deviance: 187100 > Residual Deviance: 52000 AIC: 52010 > [1] "Start reading genotype plink file here" > nbyte: 38183 > nbyte: 33735 > reserve: 360007520 > > M: 10671, N: 152732 > size of genoVecofPointers: 1 > here > time: 47915 > [1] "Genotype reading is done" > inital tau is 1 0.5 > iGet_Coef: 1 > iter from getPCG1ofSigmaAndVector 1 > iter from getPCG1ofSigmaAndVector 1 > iter from getPCG1ofSigmaAndVector 1 > Tau: > [1] 1.0 0.5 > Fixed-effect coefficients: > [,1] > [1,] NaN > [2,] NaN > Error in if (max(abs(alpha - alpha0)/(abs(alpha) + abs(alpha0) + > tol.coef)) < : > missing value where TRUE/FALSE needed > Timing stopped at: 73.18 9.298 52.25 > > > I am not sure what this means. > > Thanks a lot for all the help. > > Nicola > > > > Nicola Pirastu > Chancellor’s Fellow. > Usher Institute > The University of Edinburgh > Room 3.732, Doorway 1 > Old Medical School > Teviot Place > Edinburgh EH8 9AG > > > > Il giorno 08 mar 2018, alle ore 16:40, weizhouUMICH < > ***@***.******@***.***><mailto: ***@***.***>> ha scritto: > > Dear Nicola, > > Thank you for letting me know! I have made a version 0.26.1, in which > the inv_sympd() > is replaced by inv(). > > https://www.dropbox.com/sh/pcql2hrdid63czc/ AAC7Lm4wkbtqK9zththZWCGFa?dl=0 > > Could you please try it out? > > Btw, if you don't mind sharing the error message for the compiling, I > would > take a look to see if I can fix it. > > Thanks, > Wei > > > > Wei Zhou > Ph.D. Candidate > Bioinformatic Program > University of Michigan-Ann Arbor > > On Wed, Mar 7, 2018 at 5:38 AM, npirastu ***@***.***< ***@***.***><mailto: > ***@***.******@***.***>>> wrote: > > > For this model I used one covariate which is not monomorphic if you are > > wondering. > > > > Dear Wei, > > > > I think the issue is that kinship matrices are often not exactly > positive > > definite and the inv_sympd() does not work. > > I have tried to change it to inv() and recompile the package but it > > appears that one of the third party cget package some files are not > there > > and there are links in it’s place which makes the compiling crush. > > I have run this dataset before with GCTA which worked, of course the > > probit model is horrible so I was very happy to be able to run a proper > > logit model. > > > > Many thanks > > > > Nicola > > > > > > > > > > > > > > > > Nicola Pirastu > > Chancellor’s Fellow. > > Usher Institute > > The University of Edinburgh > > Room 3.732, Doorway 1 > > Old Medical School > > Teviot Place > > Edinburgh EH8 9AG > > > > > > ***@***.***[cid: > > ***@***.***<mailto:F ***@***.***><mailto:F > ***@***.***<mailto:D ***@***.***>>] > > > > Il giorno 07 mar 2018, alle ore 04:57, weizhouUMICH < > > ***@***.******@***.***><mailto: ***@***.***><mailto: > ***@***.******@***.***>>> ha scritto: > > > > > > Dear Nicola, > > > > Thank you so much for providing more details! Sorry for dropping the > ball > > for a while. I'm happy to help. > > You mentioned that when no covariates were used, you saw the same error > > message. > > May I ask how many covariates are used for the jobs shown in the log > file? > > > > Thanks, > > Wei > > > > — > > You are receiving this because you authored the thread. > > Reply to this email directly, view it on GitHub<https://github.com/ > > weizhouUMICH/SAIGE#17#issuecomment-371022949>, or mute the > thread< > > https://github.com/notifications/unsubscribe-auth/Ai0XChbNWmvjszW8WQ- > > EdjKLO0IZwqmJks5tb2jKgaJpZM4SHBvO>. > > > > > > The University of Edinburgh is a charitable body, registered in > > Scotland, with registration number SC005336. > > > > — > > You are receiving this because you commented. > > Reply to this email directly, view it on GitHub > > <#17# issuecomment-371097200>, > > > or mute the thread > > <https://github.com/notifications/unsubscribe-auth/AJfdL7YX_U_ > dpACd26IssQiZotRcI3Wnks5tb7ibgaJpZM4SHBvO> > > . > > > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub<https://github.com/ > weizhouUMICH/SAIGE#17#issuecomment-371545068>, or mute the thread< > https://github.com/notifications/unsubscribe-auth/Ai0XCtidP15i- > s9VD4IZCOEk39FMxM1Rks5tcV8agaJpZM4SHBvO>. > > > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#17 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/ AJfdL2jGQuVAvei9BCs3JIV7XZtRRdXrks5te6x2gaJpZM4SHBvO> > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://github.com/ weizhouUMICH/SAIGE#17#issuecomment-373753746>, or mute the thread< https://github.com/notifications/unsubscribe-auth/Ai0XCtAFNB9qvqB13e- RaspPQbzyLY6_ks5te91MgaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJfdL1XcDbiXWdirRFyPBAYkozIHO0Mbks5te-AOgaJpZM4SHBvO> .

from saige.

npirastu commented on August 23, 2024

Dear Wei, here’s the new output: M: 10671, N: 99 size of genoVecofPointers: 1 here time: 45.705 [1] "Genotype reading is done" inital tau is 1 0.5 iGet_Coef: 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 re.coef$cov: NaN NaN 0 NaN re.coef$Sigma_iX: NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN X: -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -0.07974977 -0.788817 -0.4151616 1.38986 0.01953004 -0.0393048 -0.2328933 -3.209997 0.6207993 0.8942355 -0.3338062 -1.873469 0.1249467 1.33211 0.5184882 1.669455 -0.1534982 1.057121 -1.696128 0.8965702 -0.6077298 -0.09570841 -1.096816 -1.559142 -0.06841314 0.1484303 0.5139157 -0.4743558 -0.17983 -1.139673 0.7231793 0.2152666 -1.029023 -0.2215924 -0.2958078 -0.6478647 -0.4384482 0.07254195 0.6202964 -0.4080381 2.317994 1.11522 -1.025295 -0.108153 0.007951344 0.4605968 -0.7771759 1.254086 0.4339252 1.79365 -0.144825 -1.043396 -0.8638546 -0.3958679 0.5721337 1.182113 0.3987589 0.7440897 -1.116444 -1.595013 -0.09620822 0.3846312 0.306187 0.593701 -0.8289956 1.81365 0.1594957 -0.3564891 -0.1736728 -1.733334 0.08284035 0.4424332 -1.968456 0.08998163 1.071192 -1.839723 0.0383429 1.171515 -0.7542824 -0.003146338 -1.141178 1.877035 1.096546 2.321842 -0.326046 -0.6389182 0.8493898 0.5236789 1.21312 0.1767615 -0.8480948 0.6114235 -0.6352869 -1.958003 1.269004 0.4907394 0.172751 -0.8599504 0.4635506 Tau: [1] 1.0 0.5 Fixed-effect coefficients: [,1] [1,] NaN [2,] NaN Error in if (max(abs(alpha - alpha0)/(abs(alpha) + abs(alpha0) + tol.coef)) < : missing value where TRUE/FALSE needed Timing stopped at: 0.429 0.276 0.372 Best N Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG [cid:[email protected]][cid:[email protected]] Il giorno 16 mar 2018, alle ore 18:54, weizhouUMICH <[email protected]<mailto:[email protected]>> ha scritto: Hi Nicola, Thank you so much for your quick feedback! It is likely that there is the perfect separation in the model. (one of the covariate can perfectly separate 0 or 1 for the phenotype) To test this, I made the program output the "mu", which will close to 0 or 1 if there is perfect separation. Could you please have a test run again? The program version is still the 0.26.6 in the same folder. It has been just updated Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor

On Fri, Mar 16, 2018 at 11:53 AM, npirastu ***@***.******@***.***>> wrote: Hi Wei, here’s the output. I understand now why it would not work, but not sure why this happens. re.coef$cov: NaN NaN 0 NaN re.coef$Sigma_iX: NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN X: -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -0.07974977 -0.788817 -0.4151616 1.38986 0.01953004 -0.0393048 -0.2328933 -3.209997 0.6207993 0.8942355 -0.3338062 -1.873469 0.1249467 1.33211 0.5184882 1.669455 -0.1534982 1.057121 -1.696128 0.8965702 -0.6077298 -0.09570841 -1.096816 -1.559142 -0.06841314 0.1484303 0.5139157 -0.4743558 -0.17983 -1.139673 0.7231793 0.2152666 -1.029023 -0.2215924 -0.2958078 -0.6478647 -0.4384482 0.07254195 0.6202964 -0.4080381 2.317994 1.11522 -1.025295 -0.108153 0.007951344 0.4605968 -0.7771759 1.254086 0.4339252 1.79365 -0.144825 -1.043396 -0.8638546 -0.3958679 0.5721337 1.182113 0.3987589 0.7440897 -1.116444 -1.595013 -0.09620822 0.3846312 0.306187 0.593701 -0.8289956 1.81365 0.1594957 -0.3564891 -0.1736728 -1.733334 0.08284035 0.4424332 -1.968456 0.08998163 1.071192 -1.839723 0.0383429 1.171515 -0.7542824 -0.003146338 -1.141178 1.877035 1.096546 2.321842 -0.326046 -0.6389182 0.8493898 0.5236789 1.21312 0.1767615 -0.8480948 0.6114235 -0.6352869 -1.958003 1.269004 0.4907394 0.172751 -0.8599504 0.4635506 Best N Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG ***@***.***[cid: ***@***.******@***.***>] Il giorno 16 mar 2018, alle ore 15:42, weizhouUMICH < ***@***.******@***.******@***.***>> ha scritto: Hi Nicola, Thank you for letting me know! It seems the inv() still does not work correctly. I have added some command to output the "cov", "Sigma_iX", and "X" after applying the inv() function. You can find the version 0.26.6 in the folder testingVersion/forNicola https://www.dropbox.com/sh/pcql2hrdid63czc/AAC7Lm4wkbtqK9zththZWCGFa?dl=0 Could you please run your job using this script and let me know how these three variables look? Thanks, Wei Wei Zhou Ph.D. Candidate Bioinformatic Program University of Michigan-Ann Arbor On Fri, Mar 16, 2018 at 8:13 AM, npirastu ***@***.******@***.***>< ***@***.***>> wrote: > Dear Wei, > > sorry for taking so long to respond. > > The problem with the inversion of the matrix seems to be fixed. > I am now getting: > > > CVD is a binary trait > > Call: glm(formula = formula.new, family = binomial, data = data.new) > > Coefficients: > minus1 Coronary.heart.disease_8 > 3.0067 0.2673 > > Degrees of Freedom: 134938 Total (i.e. Null); 134936 Residual > Null Deviance: 187100 > Residual Deviance: 52000 AIC: 52010 > [1] "Start reading genotype plink file here" > nbyte: 38183 > nbyte: 33735 > reserve: 360007520 > > M: 10671, N: 152732 > size of genoVecofPointers: 1 > here > time: 47915 > [1] "Genotype reading is done" > inital tau is 1 0.5 > iGet_Coef: 1 > iter from getPCG1ofSigmaAndVector 1 > iter from getPCG1ofSigmaAndVector 1 > iter from getPCG1ofSigmaAndVector 1 > Tau: > [1] 1.0 0.5 > Fixed-effect coefficients: > [,1] > [1,] NaN > [2,] NaN > Error in if (max(abs(alpha - alpha0)/(abs(alpha) + abs(alpha0) + > tol.coef)) < : > missing value where TRUE/FALSE needed > Timing stopped at: 73.18 9.298 52.25 > > > I am not sure what this means. > > Thanks a lot for all the help. > > Nicola > > > > Nicola Pirastu > Chancellor’s Fellow. > Usher Institute > The University of Edinburgh > Room 3.732, Doorway 1 > Old Medical School > Teviot Place > Edinburgh EH8 9AG > > > > Il giorno 08 mar 2018, alle ore 16:40, weizhouUMICH < > ***@***.******@***.******@***.***><mailto: ***@***.******@***.***>>> ha scritto: > > Dear Nicola, > > Thank you for letting me know! I have made a version 0.26.1, in which > the inv_sympd() > is replaced by inv(). > > https://www.dropbox.com/sh/pcql2hrdid63czc/ AAC7Lm4wkbtqK9zththZWCGFa?dl=0 > > Could you please try it out? > > Btw, if you don't mind sharing the error message for the compiling, I > would > take a look to see if I can fix it. > > Thanks, > Wei > > > > Wei Zhou > Ph.D. Candidate > Bioinformatic Program > University of Michigan-Ann Arbor > > On Wed, Mar 7, 2018 at 5:38 AM, npirastu ***@***.******@***.***>< ***@***.***><mailto: > ***@***.******@***.******@***.***>>> wrote: > > > For this model I used one covariate which is not monomorphic if you are > > wondering. > > > > Dear Wei, > > > > I think the issue is that kinship matrices are often not exactly > positive > > definite and the inv_sympd() does not work. > > I have tried to change it to inv() and recompile the package but it > > appears that one of the third party cget package some files are not > there > > and there are links in it’s place which makes the compiling crush. > > I have run this dataset before with GCTA which worked, of course the > > probit model is horrible so I was very happy to be able to run a proper > > logit model. > > > > Many thanks > > > > Nicola > > > > > > > > > > > > > > > > Nicola Pirastu > > Chancellor’s Fellow. > > Usher Institute > > The University of Edinburgh > > Room 3.732, Doorway 1 > > Old Medical School > > Teviot Place > > Edinburgh EH8 9AG > > > > > > ***@***.***[cid: > > ***@***.******@***.***><mailto:F ***@***.******@***.***>><mailto:F > ***@***.******@***.***><mailto:D ***@***.******@***.***>>>] > > > > Il giorno 07 mar 2018, alle ore 04:57, weizhouUMICH < > > ***@***.******@***.******@***.***><mailto: ***@***.******@***.***>><mailto: > ***@***.******@***.******@***.***>>> ha scritto: > > > > > > Dear Nicola, > > > > Thank you so much for providing more details! Sorry for dropping the > ball > > for a while. I'm happy to help. > > You mentioned that when no covariates were used, you saw the same error > > message. > > May I ask how many covariates are used for the jobs shown in the log > file? > > > > Thanks, > > Wei > > > > — > > You are receiving this because you authored the thread. > > Reply to this email directly, view it on GitHub<https://github.com/ > > weizhouUMICH/SAIGE#17#issuecomment-371022949>, or mute the > thread< > > https://github.com/notifications/unsubscribe-auth/Ai0XChbNWmvjszW8WQ- > > EdjKLO0IZwqmJks5tb2jKgaJpZM4SHBvO>. > > > > > > The University of Edinburgh is a charitable body, registered in > > Scotland, with registration number SC005336. > > > > — > > You are receiving this because you commented. > > Reply to this email directly, view it on GitHub > > <#17# issuecomment-371097200>, > > > or mute the thread > > <https://github.com/notifications/unsubscribe-auth/AJfdL7YX_U_ > dpACd26IssQiZotRcI3Wnks5tb7ibgaJpZM4SHBvO> > > . > > > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub<https://github.com/ > weizhouUMICH/SAIGE#17#issuecomment-371545068>, or mute the thread< > https://github.com/notifications/unsubscribe-auth/Ai0XCtidP15i- > s9VD4IZCOEk39FMxM1Rks5tcV8agaJpZM4SHBvO>. > > > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#17 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/ AJfdL2jGQuVAvei9BCs3JIV7XZtRRdXrks5te6x2gaJpZM4SHBvO> > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://github.com/ weizhouUMICH/SAIGE#17#issuecomment-373753746>, or mute the thread< https://github.com/notifications/unsubscribe-auth/Ai0XCtAFNB9qvqB13e- RaspPQbzyLY6_ks5te91MgaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AJfdL1XcDbiXWdirRFyPBAYkozIHO0Mbks5te-AOgaJpZM4SHBvO> .

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XCiz1QZdoXoBgKXF4sGgkJHUxiQasks5tfAp-gaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

weizhouUMICH commented on August 23, 2024

Hi Nicola, Thanks! Sorry I just realized the version 0.26.6 was not updated successfully, so the output you got is exactly the same as the previous one. I just updated it. Could you please try again? Thanks, Wei

from saige.

npirastu commented on August 23, 2024

Hi Wei, Here’s the output. As you can see they are not all 0 or 1. I was wondering if the problem could not be in the kinship matrix estimation since all values of Sigma_Xi are NaN. Thanks a lot N M: 10671, N: 99 size of genoVecofPointers: 1 here time: 67.943 [1] "Genotype reading is done" mu: 0.05845516 0.05044995 0.05452984 0.07896694 0.05966697 0.05894602 0.05663108 0.03027144 0.06752222 0.07140339 0.05545849 0.04018498 0.06097941 0.07804928 0.06612098 0.0835506 0.05756997 0.07381229 0.04171459 0.0714374 0.0523898 0.05826253 0.04730596 0.04293409 0.05859236 0.06127544 0.06605899 0.05386338 0.05725697 0.04688352 0.06895195 0.06212533 0.04798159 0.05676383 0.05589732 0.05195388 0.05426674 0.06032363 0.06751527 0.05461056 0.09512502 0.07468945 0.04801901 0.05811274 0.05952444 0.06534012 0.05057258 0.07682498 0.06498325 0.08566323 0.05767341 0.04783758 0.04966622 0.05474874 0.0668523 0.07571126 0.0645155 0.06924738 0.04711204 0.0426115 0.05825651 0.06432847 0.06329914 0.06714843 0.05002886 0.08600793 0.06141539 0.05519809 0.05733002 0.04138913 0.06045198 0.06509689 0.03938795 0.06054113 0.07402384 0.04047186 0.05989923 0.07554851 0.05081457 0.05938812 0.04686875 0.08710869 0.07440647 0.09519779 0.05554785 0.05205075 0.07075302 0.06619141 0.07618923 0.06163436 0.04982986 0.06739268 0.05209012 0.03947492 0.07705769 0.06574561 0.06158344 0.04970671 0.06537975 inital tau is 1 0.5 iGet_Coef: 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 iter from getPCG1ofSigmaAndVector 1 Tau: [1] 1.0 0.5 Fixed-effect coefficients: [,1] [1,] NaN [2,] NaN Error in if (max(abs(alpha - alpha0)/(abs(alpha) + abs(alpha0) + tol.coef)) < : missing value where TRUE/FALSE needed Timing stopped at: 0.431 0.127 0.395 Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG [cid:[email protected]][cid:[email protected]] Il giorno 17 mar 2018, alle ore 18:45, weizhouUMICH <[email protected]<mailto:[email protected]>> ha scritto: Hi Nicola, Thanks! Sorry I just realized the version 0.26.6 was not updated successfully, so the output you got is exactly the same as the previous one. I just updated it. Could you please try again? Thanks, Wei — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XCiUdDz5b9HOkUjq1ogl_dPg3HCwqks5tfVm_gaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

byenj commented on August 23, 2024

Hi Nicola and Wei,

Hopefully it's already solved, but just in case, other people like me met the same problem. It turns out the variants need to be filtered to at least MAC=1 for any individuals that have phenotype data for the particular trait. It's probably a good idea to include this in QC steps.

Best,
Bin

from saige.

npirastu commented on August 23, 2024

Dear Bin, thanks a lot for the info, however this has not solved the problem. I am slightly concerned because it runs extremely fast which would make me think that it is not actually reading the genotypes or estimating the kinship matrix. Not sure what I could change the error message is still the same. Best Nicola Nicola Pirastu Chancellor’s Fellow. Usher Institute The University of Edinburgh Room 3.732, Doorway 1 Old Medical School Teviot Place Edinburgh EH8 9AG [cid:[email protected]][cid:[email protected]] Il giorno 09 apr 2018, alle ore 20:42, byenj <[email protected]<mailto:[email protected]>> ha scritto: Hi Nicola and Wei, Hopefully it's already solved, but just in case, other people like me met the same problem. It turns out the variants need to be filtered to at least MAC=1 for any individuals that have phenotype data for the particular trait. It's probably a good idea to include this in QC steps. Best, Bin — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/Ai0XChPp9ZFduKVeXSql2uqEpY9TbiPNks5tm7m3gaJpZM4SHBvO>. The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

from saige.

summike commented on August 23, 2024

Dear Wei, Nicola and Bin,

I hope this problem has been solved. I am sorry that I met a similar issue with my analysis of UK Biobank data. Please kindly see the error in the attached log.

Please kindly note I am using the SAIGE under v 0.29.5. I have also checked my plink file which contains only autosomal variants whose minimum MAF is 1.00e-05.

Any comment or suggestion is appreciated.

Thank you very much!

Best wishes,
Xiao
SAIGE_bi_HPTN_1.log

from saige.

byenj commented on August 23, 2024

Hi Xiao, As far as I remembered, there're two main issues: 1. Use MAC(minor allele count) 3 or 4 filter, not MAF; 2. The phenotype need to be filtered with the same samples, i.e. all phenotyped samples need to have genotype data, and vice versa. Hope this helps! Good luck! Bin

…

On Sun, May 12, 2019, 1:09 PM summike ***@***.***> wrote: Dear Wei, Nicola and Bin, I hope this problem has been solved. I am sorry that I met a similar issue with my analysis of UK Biobank data. Please kindly see the error in the attached log. Please kindly note I am using the SAIGE under v 0.29.5. I have also checked my plink file which contains only autosomal variants whose minimum MAF is 1.00e-05. Any comment or suggestion is appreciated. Thank you very much! Best wishes, Xiao SAIGE_bi_HPTN_1.log <https://github.com/weizhouUMICH/SAIGE/files/3170352/SAIGE_bi_HPTN_1.log> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACH77JRGDQR37SD7LV55XCLPVBFLRANCNFSM4EQ4DPHA> .

from saige.

summike commented on August 23, 2024

@byenj
Hi Bin,

Thank you very much for your reply! However, even I followed your suggestions (MAC=3; only keep samples having both genotype and phenotype data), I still got the same error, but the 'instance' seems changed. Please kindly see the details below.

error: inv_sympd(): matrix is singular or not positive definite
terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc

compared with last time:

error: inv_sympd(): matrix is singular or not positive definite
terminate called after throwing an instance of 'std::length_error'
  what():  basic_string::_M_create

Really appreciate you helping me with this!

Many thanks and best wishes,
Xiao

from saige.

weizhouUMICH commented on August 23, 2024

Hi Xiao,

This issue should have been solved in more recent versions.

Thanks,
Wei

from saige.

UK biobank analysis about saige HOT 21 CLOSED

Comments (21)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent