I have the following problem: I measured co2 on 6 paired sites (one grubbed and one non-grubbed fence per site -> grubbing = treatment). These measurements I repeated 15 times over 2 years. So, now my problem is how to analyze these data. I tried the following model: mod1_CO2<-lme(co2~treatment+time,random=~1|site,data=CO2_t1_t15) I think with the random effect I included the paired sample design of my study, but obviously the model does not check for the "repeated measurements" structure of my data. Or in other words, the model assumpt that different sampling dates are independent of each other, but they aren't (since I measured the CO2 15 times on exactly the same spot). So, that's why my question is how can I combine the paired (or nested) design with the repeated measurements in one model??? Thank you for any help Sven -- Sven Wirthner Eidg. Forschungsanstalt f?r Wald, Schnee und Landschaft WSL Tier?kologie Z?rcherstrasse 111 CH-8903 Birmensdorf Tel. +41 44 7392 371 Fax +41 44 7392 215 sven.wirthner at wsl.ch http://www.wsl.ch ----------------------------------------------------------------------------------- This message was sent using IMP (http://horde.org/imp/) at WSL (http://www.wsl.ch).
Sven: 1. Try this question on the r-sig-mixwed-models list -- more likely to get good responses there, I think. 2. However, my less expert opinion is that you need a random fence within site term (each fence has it's own ID) co2 ~ treatment + time, rand = ~1|fence/site This assumes time is continuous and not a factor, btw; if time is a factor, I think there may be identifiability problems -- but folks on the mixed models list should be able to provide you better guidance. There may well be better ways to model the covariance structure. Bert Gunter Genentech Nonclinical Biostatistics -----Original Message----- From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf Of Sven Wirthner Sent: Thursday, January 14, 2010 5:24 AM To: r-help at r-project.org Subject: [R] paired repeated measurements I have the following problem: I measured co2 on 6 paired sites (one grubbed and one non-grubbed fence per site -> grubbing = treatment). These measurements I repeated 15 times over 2 years. So, now my problem is how to analyze these data. I tried the following model: mod1_CO2<-lme(co2~treatment+time,random=~1|site,data=CO2_t1_t15) I think with the random effect I included the paired sample design of my study, but obviously the model does not check for the "repeated measurements" structure of my data. Or in other words, the model assumpt that different sampling dates are independent of each other, but they aren't (since I measured the CO2 15 times on exactly the same spot). So, that's why my question is how can I combine the paired (or nested) design with the repeated measurements in one model??? Thank you for any help Sven -- Sven Wirthner Eidg. Forschungsanstalt f?r Wald, Schnee und Landschaft WSL Tier?kologie Z?rcherstrasse 111 CH-8903 Birmensdorf Tel. +41 44 7392 371 Fax +41 44 7392 215 sven.wirthner at wsl.ch http://www.wsl.ch ---------------------------------------------------------------------------- ------- This message was sent using IMP (http://horde.org/imp/) at WSL (http://www.wsl.ch). ______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Just realized that I got my nesting specification backwards. For fence in site, which is what I wanted, it should have been: co2 ~ treatment + time, rand = ~1|site/fence Bert Gunter Genentech Nonclinical Biostatistics -----Original Message----- From: bgunter Sent: Thursday, January 14, 2010 10:59 AM To: Sven Wirthner; r-help at r-project.org Subject: RE: [R] paired repeated measurements Sven: 1. Try this question on the r-sig-mixwed-models list -- more likely to get good responses there, I think. 2. However, my less expert opinion is that you need a random fence within site term (each fence has it's own ID) co2 ~ treatment + time, rand = ~1|fence/site This assumes time is continuous and not a factor, btw; if time is a factor, I think there may be identifiability problems -- but folks on the mixed models list should be able to provide you better guidance. There may well be better ways to model the covariance structure. Bert Gunter Genentech Nonclinical Biostatistics -----Original Message----- From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf Of Sven Wirthner Sent: Thursday, January 14, 2010 5:24 AM To: r-help at r-project.org Subject: [R] paired repeated measurements I have the following problem: I measured co2 on 6 paired sites (one grubbed and one non-grubbed fence per site -> grubbing = treatment). These measurements I repeated 15 times over 2 years. So, now my problem is how to analyze these data. I tried the following model: mod1_CO2<-lme(co2~treatment+time,random=~1|site,data=CO2_t1_t15) I think with the random effect I included the paired sample design of my study, but obviously the model does not check for the "repeated measurements" structure of my data. Or in other words, the model assumpt that different sampling dates are independent of each other, but they aren't (since I measured the CO2 15 times on exactly the same spot). So, that's why my question is how can I combine the paired (or nested) design with the repeated measurements in one model??? Thank you for any help Sven -- Sven Wirthner Eidg. Forschungsanstalt f?r Wald, Schnee und Landschaft WSL Tier?kologie Z?rcherstrasse 111 CH-8903 Birmensdorf Tel. +41 44 7392 371 Fax +41 44 7392 215 sven.wirthner at wsl.ch http://www.wsl.ch ---------------------------------------------------------------------------- ------- This message was sent using IMP (http://horde.org/imp/) at WSL (http://www.wsl.ch). ______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.