Hello my programmer friends. I am wondering if anyone has written a corStruct for this lme call: lme(y~time*group, random= ~time|id) In this call, the covariance structure implied by (time|id) is assumed constant across groups. We would like to model these random effects separately in the two groups. Motivation: In a clinical trail with an _effective_ treatment, response to treatment will be variable (variance of the random slopes will be greater in the treatment arm). Thank you in advance for any thoughts where I might find such a corStruct. Steve Edland & Yu Zhao Sincerely, Steve Edland Steven D. Edland, Ph.D. Professor Dept. of Family Medicine & Public Health Dept. of Neurosciences University of California, San Diego 9500 Gilman Dr. M/C 0948 La Jolla, CA 92093-0948 http://biostat.ucsd.edu/sedland.htm
novel correlation structure for nlme package, lme function
4 messages · Thierry Onkelinx, Edland, Steven
Dear Steve, I'd create two variables time_treatment <- ifelse(group == "treatment", time, 0) time_placebo <- ifelse(group != "treatment", time, 0) And then use ~ time_treatment + time_placebo | id as random effect. This should give you the two random slope, each with it own variance. Best regards, ir. Thierry Onkelinx Statisticus / Statistician Vlaamse Overheid / Government of Flanders INSTITUUT VOOR NATUUR- EN BOSONDERZOEK / RESEARCH INSTITUTE FOR NATURE AND FOREST Team Biometrie & Kwaliteitszorg / Team Biometrics & Quality Assurance thierry.onkelinx at inbo.be Havenlaan 88 bus 73, 1000 Brussel www.inbo.be /////////////////////////////////////////////////////////////////////////////////////////// To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher The plural of anecdote is not data. ~ Roger Brinner The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data. ~ John Tukey /////////////////////////////////////////////////////////////////////////////////////////// <https://www.inbo.be> Op do 16 jul. 2020 om 20:35 schreef Edland, Steven <sedland at health.ucsd.edu
:
Hello my programmer friends. I am wondering if anyone has written a corStruct for this lme call: lme(y~time*group, random= ~time|id) In this call, the covariance structure implied by (time|id) is assumed constant across groups. We would like to model these random effects separately in the two groups. Motivation: In a clinical trail with an _effective_ treatment, response to treatment will be variable (variance of the random slopes will be greater in the treatment arm). Thank you in advance for any thoughts where I might find such a corStruct. Steve Edland & Yu Zhao Sincerely, Steve Edland Steven D. Edland, Ph.D. Professor Dept. of Family Medicine & Public Health Dept. of Neurosciences University of California, San Diego 9500 Gilman Dr. M/C 0948 La Jolla, CA 92093-0948 http://biostat.ucsd.edu/sedland.htm [[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
Thanks Thierry for your reply! We did try that, and it does not work as best as we can tell. It becomes a problem when the allocation ratio is not 1:1. E.g., trials often use 2:1 treatment:control allocation ratio. In this case the s.e. of the fixed effects slope coefficients are off, and we do not meet the nominal type I error rate. Any experiences you can share would be appreciate, and thank you again for replying. Regards, Steve Steven D. Edland, Ph.D. Professor Dept. of Family Medicine & Public Health Dept. of Neurosciences University of California, San Diego 9500 Gilman Dr. M/C 0948 La Jolla, CA 92093-0948 http://biostat.ucsd.edu/sedland.htm
From: Thierry Onkelinx <thierry.onkelinx at inbo.be>
Sent: Thursday, July 16, 2020 11:53 AM
To: Edland, Steven <sedland at health.ucsd.edu>
Cc: r-sig-mixed-models at r-project.org <r-sig-mixed-models at r-project.org>; Yu Zhao <yuz867 at ucsd.edu>
Subject: Re: [R-sig-ME] novel correlation structure for nlme package, lme function
Sent: Thursday, July 16, 2020 11:53 AM
To: Edland, Steven <sedland at health.ucsd.edu>
Cc: r-sig-mixed-models at r-project.org <r-sig-mixed-models at r-project.org>; Yu Zhao <yuz867 at ucsd.edu>
Subject: Re: [R-sig-ME] novel correlation structure for nlme package, lme function
Dear Steve, I'd create two variables time_treatment <- ifelse(group == "treatment", time, 0) time_placebo <- ifelse(group != "treatment", time, 0) And then use ~ time_treatment + time_placebo | id as random effect. This should give you the two random slope, each with it own variance. Best regards, ir. Thierry Onkelinx Statisticus / Statistician Vlaamse Overheid / Government of Flanders INSTITUUT VOOR NATUUR- EN BOSONDERZOEK / RESEARCH INSTITUTE FOR NATURE AND FOREST Team Biometrie & Kwaliteitszorg / Team Biometrics & Quality Assurance thierry.onkelinx at inbo.be<mailto:thierry.onkelinx at inbo.be> Havenlaan 88 bus 73, 1000 Brussel www.inbo.be<http://www.inbo.be> /////////////////////////////////////////////////////////////////////////////////////////// To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher The plural of anecdote is not data. ~ Roger Brinner The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data. ~ John Tukey /////////////////////////////////////////////////////////////////////////////////////////// [https://inbo-website-prd-532750756126.s3-eu-west-1.amazonaws.com/inbologoleeuw_nl.png]<https://www.inbo.be> Op do 16 jul. 2020 om 20:35 schreef Edland, Steven <sedland at health.ucsd.edu<mailto:sedland at health.ucsd.edu>>: Hello my programmer friends. I am wondering if anyone has written a corStruct for this lme call: lme(y~time*group, random= ~time|id) In this call, the covariance structure implied by (time|id) is assumed constant across groups. We would like to model these random effects separately in the two groups. Motivation: In a clinical trail with an _effective_ treatment, response to treatment will be variable (variance of the random slopes will be greater in the treatment arm). Thank you in advance for any thoughts where I might find such a corStruct. Steve Edland & Yu Zhao Sincerely, Steve Edland Steven D. Edland, Ph.D. Professor Dept. of Family Medicine & Public Health Dept. of Neurosciences University of California, San Diego 9500 Gilman Dr. M/C 0948 La Jolla, CA 92093-0948 http://biostat.ucsd.edu/sedland.htm [[alternative HTML version deleted]] _______________________________________________ R-sig-mixed-models at r-project.org<mailto:R-sig-mixed-models at r-project.org> mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
Hi Thierry, Apologies, I read your email too quickly. This syntax for specifying the covariance structure works well for my application. Thank you again, and sincere apologies for my hasty reply below. Regards, Steve Steven D. Edland, Ph.D. Professor Dept. of Family Medicine & Public Health Dept. of Neurosciences University of California, San Diego 9500 Gilman Dr. M/C 0948 La Jolla, CA 92093-0948 http://biostat.ucsd.edu/sedland.htm
From: Edland, Steven <sedland at health.ucsd.edu>
Sent: Thursday, July 16, 2020 12:51 PM
To: Thierry Onkelinx <thierry.onkelinx at inbo.be>
Cc: r-sig-mixed-models at r-project.org <r-sig-mixed-models at r-project.org>; Yu Zhao <yuz867 at ucsd.edu>
Subject: Re: [R-sig-ME] novel correlation structure for nlme package, lme function
Sent: Thursday, July 16, 2020 12:51 PM
To: Thierry Onkelinx <thierry.onkelinx at inbo.be>
Cc: r-sig-mixed-models at r-project.org <r-sig-mixed-models at r-project.org>; Yu Zhao <yuz867 at ucsd.edu>
Subject: Re: [R-sig-ME] novel correlation structure for nlme package, lme function
Thanks Thierry for your reply! We did try that, and it does not work as best as we can tell. It becomes a problem when the allocation ratio is not 1:1. E.g., trials often use 2:1 treatment:control allocation ratio. In this case the s.e. of the fixed effects slope coefficients are off, and we do not meet the nominal type I error rate. Any experiences you can share would be appreciate, and thank you again for replying. Regards, Steve Steven D. Edland, Ph.D. Professor Dept. of Family Medicine & Public Health Dept. of Neurosciences University of California, San Diego 9500 Gilman Dr. M/C 0948 La Jolla, CA 92093-0948 http://biostat.ucsd.edu/sedland.htm ________________________________ From: Thierry Onkelinx <thierry.onkelinx at inbo.be> Sent: Thursday, July 16, 2020 11:53 AM To: Edland, Steven <sedland at health.ucsd.edu> Cc: r-sig-mixed-models at r-project.org <r-sig-mixed-models at r-project.org>; Yu Zhao <yuz867 at ucsd.edu> Subject: Re: [R-sig-ME] novel correlation structure for nlme package, lme function Dear Steve, I'd create two variables time_treatment <- ifelse(group == "treatment", time, 0) time_placebo <- ifelse(group != "treatment", time, 0) And then use ~ time_treatment + time_placebo | id as random effect. This should give you the two random slope, each with it own variance. Best regards, ir. Thierry Onkelinx Statisticus / Statistician Vlaamse Overheid / Government of Flanders INSTITUUT VOOR NATUUR- EN BOSONDERZOEK / RESEARCH INSTITUTE FOR NATURE AND FOREST Team Biometrie & Kwaliteitszorg / Team Biometrics & Quality Assurance thierry.onkelinx at inbo.be<mailto:thierry.onkelinx at inbo.be> Havenlaan 88 bus 73, 1000 Brussel www.inbo.be<http://www.inbo.be> /////////////////////////////////////////////////////////////////////////////////////////// To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher The plural of anecdote is not data. ~ Roger Brinner The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data. ~ John Tukey /////////////////////////////////////////////////////////////////////////////////////////// [https://inbo-website-prd-532750756126.s3-eu-west-1.amazonaws.com/inbologoleeuw_nl.png]<https://www.inbo.be> Op do 16 jul. 2020 om 20:35 schreef Edland, Steven <sedland at health.ucsd.edu<mailto:sedland at health.ucsd.edu>>: Hello my programmer friends. I am wondering if anyone has written a corStruct for this lme call: lme(y~time*group, random= ~time|id) In this call, the covariance structure implied by (time|id) is assumed constant across groups. We would like to model these random effects separately in the two groups. Motivation: In a clinical trail with an _effective_ treatment, response to treatment will be variable (variance of the random slopes will be greater in the treatment arm). Thank you in advance for any thoughts where I might find such a corStruct. Steve Edland & Yu Zhao Sincerely, Steve Edland Steven D. Edland, Ph.D. Professor Dept. of Family Medicine & Public Health Dept. of Neurosciences University of California, San Diego 9500 Gilman Dr. M/C 0948 La Jolla, CA 92093-0948 http://biostat.ucsd.edu/sedland.htm [[alternative HTML version deleted]] _______________________________________________ R-sig-mixed-models at r-project.org<mailto:R-sig-mixed-models at r-project.org> mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models