bbolker
commented
2 months ago We need a little bit more information.
library(lme4)
fitfun <- function(form) {
set.seed(101)
dd <- data.frame(x = rnorm(1000), f = factor(rep(1:20, each = 50)))
dd$y <- simulate( ~ x + (1|f), newdata = dd, newparams = list(beta = c(1,1), theta = 1, sigma = 1), family = gaussian)[[1]]
fit <- lmer (form, data = dd)
return(fit)
}
## construct formula *outside* of environment where data exist
form <- y ~ x + (1|f)
m <- fitfun(form)
library(lattice)
qqmath(m) ## object 'dd' not foundI could provoke your error if I use the name data instead of dd for the data frame generated inside the function (this way, rather than a data frame not being found, the function will find the built-in function data() ...)
The fundamental problem is that R model objects don't typically carry around a full copy of the data; they do carry a model frame, but that doesn't necessarily have the original variables in it (for example if a model uses log(x), the model frame will have log(x) in it rather than x ...
We could certainly (1) try harder to find the data (maybe looking in the model frame in case it does have what we need), (2) try harder not to need the data (e.g. qqmath(m) shouldn't need to access the data; plot.merMod is a little bit smarter, it only tries to get the data if the formula refers to a variable that needs to be drawn from the data [e.g. plot(m, residuals(.) ~ fitted(.)) works fine in the above example, we only run into trouble when try plot(m, residuals(.) ~ x).
Can you say a little bit more about where the data and formula are constructed in your workflow so we can see if there's a way to make this work for you?
tamelung
What I did.
I computed GLMMs on a high performance computing cluster. I imported these in my local R-Studio as RDS-files. Now, I am trying to plot diagnostic plots following the procedures laid out in Mixed model lab #1. I run the following code:
What happened.
I get these error messages:
What I expected to happen.
to receive diagnostic plots
potential problem / solution
in https://github.com/lme4/lme4/blob/2f37b6dcac9fea9ea81f7980e2c23a61cde73368/R/plot.R#L31 getData.merMod seems to try and get the data from the environment. In my special case using external computation power, this doesn't work. After I read in the data with the exact same name as used in the original call the plots seem fine.
sessionInfo()
R-Studio (Version 2024.04.2+764 (2024.04.2+764))
R version 4.3.3 (2024-02-29) Platform: x86_64-apple-darwin20 (64-bit) Running under: macOS Sonoma 14.4.1
Matrix products: default BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib LAPACK: /Library/Frameworks/R.framework/Versions/4.3-x86_64/Resources/lib/libRlapack.dylib; LAPACK version 3.11.0
attached base packages: [1] stats graphics grDevices utils datasets methods base
other attached packages: [1] lme4_1.1-35.3 lattice_0.22-6 interactions_1.1.5 performance_0.12.2 effectsize_0.8.8
[6] emmeans_1.10.1 data.table_1.15.4 bruceR_2024.6 gridExtra_2.3 broom.mixed_0.2.9.5 [11] tibble_3.2.1 moments_0.14.1 multcomp_1.4-25 TH.data_1.1-2 MASS_7.3-60.0.1
[16] survival_3.6-4 mvtnorm_1.2-4 tidyr_1.3.1 stringr_1.5.1 readxl_1.4.3
[21] dplyr_1.1.4 ggplot2_3.5.1 Matrix_1.6-5