jslefche
commented
5 years ago Hi, it’s a bit hard to tell what’s happening here. It could be pbkrtest trying to get p-values from lmer models or something in our package.
Try installing the development version of piecewise and see if that fixes the problem:
install.packages(“devtools”)
library(devtools)
devtools::install_github(“jslefche/piecewiseSEM@devel”)Feel free to contact me by email if you continue to run into problems
J
Jonathan S. Lefcheck, Ph.D. Tennenbaum Coordinating Scientist MarineGEO: https://marinegeo.si.edu/ Smithsonian Institution Phone: +1 (443) 482-2443 www.jonlefcheck.nethttp://www.jonlefcheck.net
From: jlamitmailto:notifications@github.com Sent: Saturday, March 23, 2019 1:27 PM To: jslefche/piecewiseSEMmailto:piecewiseSEM@noreply.github.com Cc: Subscribedmailto:subscribed@noreply.github.com Subject: [jslefche/piecewiseSEM] Errors using a random effect (#182)
Hi, I am receiving some errors that I don't quite understand. The data involves bark traits, and lichen data collected on trees in a common garden. There are 54 trees in the dataset, each tree belongs to a single tree genotype, there are 13 tree genotypes total, and each genotype is represented by 2-6 trees. On each tree the following variables were quantified: total lichen cover (TLCov) in quadrats on tree trunks, bark texture roughness on the tree trunks (BR), and condensed tannins of the bark of each tree trunk. All the three measured variables vary significantly among tree genotypes (done using a simple model with genotype as a random effect and a single trait as a response variable, in lmerTest). What I am interested in using SEM for is testing if condensed tannins or roughness are genetically based mechanisms that cause lichen cover to vary among tree genotypes. So, the simple SEM would have a direct path from tree genotype to condensed tannins and from tree genotype to bark roughness, and then genotype, condensed tannins and roughness would have paths leading to lichen cover. This seems like a relatively simple model to run but I get error messages when I fit it with lmerTest and with nlme. Below is the output from running the models using both programs. I am not certain that I understand the error messages entirely, or at least the reason for them. Any suggestions on where to go from here would be greatly appreciated. Thanks.
###################################
Below is the model set fit with lmerTest, which reveals some issues. First, it seems to think there are NAs in the dataset but there are no missing values in the input data. Second, output from looking at the Model_lmerTest object shows that no model actually has tree Genotype as an effect, and the data says 244 rows and a bunch of NAs, neither of which reflect the input dataset. Third, running summary and coefs both say there is an incorrect number of dimensions.
SEM 1 with lmerTest
Model_lmerTest<-psem(
- lmer(BR~(1|Geno),data=RawData),
- lmer(CT~(1|Geno),data=RawData),
- lmer(TLCov~BR+CT+(1|Geno),data=RawData)) Warning message: NAs detected in the dataset. Models will run but this is not recommended
Model_lmerTest Structural Equations: lmerModLmerTest: BR ~ 1 lmerModLmerTest: CT ~ 1 lmerModLmerTest: TLCov ~ BR + CT
Data: Geno BR CT TLCov 1 10 5.5 2.66 3.25 2 10 5.5 3.04 NA 3 10 5.5 3.07 NA 4 10 5.5 3.38 NA 5 10 10.0 2.66 NA 6 10 10.0 3.04 NA ...with 244 more rows
[1] "class(psem)"
summary(Model_lmerTest) |====================================================================================================================================================================================| 100% Error in newdata[, notrans[k]] : incorrect number of dimensions rsquared(Model_lmerTest, method = NULL) Response family link method Marginal Conditional 1 BR gaussian identity none 0.0000000 0.33483307 2 CT gaussian identity none 0.0000000 0.09397688 3 TLCov gaussian identity none 0.1261363 0.21848890 coefs(Model_lmerTest) Error in newdata[, notrans[k]] : incorrect number of dimensions
###################################
Below is the model set fit with nlme, which also reveals some issues. First, output from looking at the Model_nlme object shows that no model actually has tree Genotype as an effect, but unlike the lmerTest output (above) the row number is correct and there are no NAs indicated. Running summary and coefs both say there is less than 2 dimensions. The rsquared function has the same output as that for the lmerTest model set.
Model_nlme <- psem(
- lme(BR ~1, random = ~1|Geno, data=RawData),
- lme(CT~1, random = ~1|Geno, data=RawData),
- lme(TLCov~ BR +CT, random = ~1|Geno, data=RawData))
Model_nlme Structural Equations: lme: BR ~ 1 lme: CT ~ 1 lme: TLCov ~ BR + CT
Data: Tree.ID Geno BR TLCov CT 1 N1.04 1017 75.0 15.13 2.77 2 N1.07 T6 40.0 5.63 1.93 3 N1.08 1017 37.5 3.05 2.07 4 N1.10 996 17.5 2.25 2.43 5 N1.11 T6 85.0 25.03 2.46 6 N1.24 1017 55.0 32.25 1.75 ...with 54 more rows
[1] "class(psem)"
summary(Model_nlme) |====================================================================================================================================================================================| 100% Error in colnames<-(tmp, value = notrans) : attempt to set 'colnames' on an object with less than two dimensions rsquared(Model_nlme, method = NULL) Response family link method Marginal Conditional 1 BR gaussian identity none 0.0000000 0.33483306 2 CT gaussian identity none 0.0000000 0.09397688 3 TLCov gaussian identity none 0.1261362 0.21848891 coefs(Model_nlme) Error in colnames<-(tmp, value = notrans) : attempt to set 'colnames' on an object with less than two dimensions
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fjslefche%2FpiecewiseSEM%2Fissues%2F182&data=02%7C01%7Clefcheckj%40si.edu%7C8ce66f38909a4c466ed308d6afb4c7b3%7C989b5e2a14e44efe93b78cdd5fc5d11c%7C1%7C0%7C636889588315332200&sdata=j9z5qh1AN63PfY34xe%2BzIMjwEGO1N4koLW1GCzWoM50%3D&reserved=0, or mute the threadhttps://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FACPAVwaxxsZX7RO39ONRbu87xTCDteTeks5vZmPsgaJpZM4cFCky&data=02%7C01%7Clefcheckj%40si.edu%7C8ce66f38909a4c466ed308d6afb4c7b3%7C989b5e2a14e44efe93b78cdd5fc5d11c%7C1%7C0%7C636889588315342200&sdata=%2F8p6w%2Bx6fn2lrmiRawwCuM48FJ6q17eFZGI2IgAYXzI%3D&reserved=0.
MKLau
rachelthoms
jebyrnes
Hi, I am receiving some errors that I don't quite understand. The data involves bark traits, and lichen data collected on trees in a common garden. There are 54 trees in the dataset, each tree belongs to a single tree genotype, there are 13 tree genotypes total, and each genotype is represented by 2-6 trees. On each tree the following variables were quantified: total lichen cover (TLCov) in quadrats on tree trunks, bark texture roughness on the tree trunks (BR), and condensed tannins of the bark of each tree trunk. All the three measured variables vary significantly among tree genotypes (done using a simple model with genotype as a random effect and a single trait as a response variable, in lmerTest). What I am interested in using SEM for is testing if condensed tannins or roughness are genetically based mechanisms that cause lichen cover to vary among tree genotypes. So, the simple SEM would have a direct path from tree genotype to condensed tannins and from tree genotype to bark roughness, and then genotype, condensed tannins and roughness would have paths leading to lichen cover. This seems like a relatively simple model to run but I get error messages when I fit it with lmerTest and with nlme. Below is the output from running the models using both programs. I am not certain that I understand the error messages entirely, or at least the reason for them. Any suggestions on where to go from here would be greatly appreciated. Thanks.
###################################
Below is the model set fit with lmerTest, which reveals some issues. First, it seems to think there are NAs in the dataset but there are no missing values in the input data. Second, output from looking at the Model_lmerTest object shows that no model actually has tree Genotype as an effect, and the data says 244 rows and a bunch of NAs, neither of which reflect the input dataset. Third, running summary and coefs both say there is an incorrect number of dimensions.
Data: Geno BR CT TLCov 1 10 5.5 2.66 3.25 2 10 5.5 3.04 NA 3 10 5.5 3.07 NA 4 10 5.5 3.38 NA 5 10 10.0 2.66 NA 6 10 10.0 3.04 NA ...with 244 more rows
[1] "class(psem)"
###################################
Below is the model set fit with nlme, which also reveals some issues. First, output from looking at the Model_nlme object shows that no model actually has tree Genotype as an effect, but unlike the lmerTest output (above) the row number is correct and there are no NAs indicated. Running summary and coefs both say there is less than 2 dimensions. The rsquared function has the same output as that for the lmerTest model set.
Data: Tree.ID Geno BR TLCov CT 1 N1.04 1017 75.0 15.13 2.77 2 N1.07 T6 40.0 5.63 1.93 3 N1.08 1017 37.5 3.05 2.07 4 N1.10 996 17.5 2.25 2.43 5 N1.11 T6 85.0 25.03 2.46 6 N1.24 1017 55.0 32.25 1.75 ...with 54 more rows
[1] "class(psem)"