It seems like there's some underlying structure issue in my data that's causing all my models to break. I tried using a linear mixed effects models in nlme, lmer, and repeated measures anovas. Every time, I run into some variation of the 'model is singular' error:
Linear Mixed Effects Model (nlme):
lme.totalmoss3\<-lme(moss_TOT\~permafrost\*year, data = veg1, random = \~ 1 | plot, method="REML", na.action = na.omit) intervals(lme.totalmoss3) Approximate 95% confidence intervals
intervals(lme.totalmoss3)
Approximate 95% confidence intervals
Fixed effects: lower est. upper
(Intercept) -1.582048 3.783333 9.148714
permafrostNOICE -10.670785 -1.928788 6.813209
year2021 14.920823 22.535000 30.149177
permafrostNOICE:year2021 -2.574414 9.788636 22.151687
Random Effects:
Level: plot
lower est. upper
sd((Intercept)) 3.564458e-55 0.001182198 3.92091e+48
\#breaking, assumed this was a power issue so switched to anovas
Within-group standard error:
lower est. upper
9.202664 11.112724 13.419226
Linear Mixed Effects Model (lmer):
lme.totalmoss3\<-lmer(moss_TOT\~permafrost\*year + (1 | plot), data = veg1, REML = TRUE, na.action = na.omit)
boundary (singular) fit: see help('isSingular')
Repeated Measures Anova:
aov.totalmoss3\<-aov(moss_TOT\~permafrost\*year+Error(plot/(permafrost\*year)), data = veg1)
Warning message: In aov(moss_TOT \~ permafrost \* year + Error(plot/permafrost \* year), : Error() model is singular
(different way): aov.totalmoss3\<-anova_test(data = veg1,
dv = moss_TOT,
wid = plot,
within = c(permafrost, year),
effect.size = 'pes')
Error in lm.fit(x, y, offset = offset, singular.ok = singular.ok, ...) : 0 (non-NA) cases
As you can see, I'm looking at the effect of permafrost presence (ice, no ice) and year (2015, 2021) on total moss % cover values. Unfortunately I can't include my datasheet since the data is confidential. But here are the column headings:
Any advice would be greatly appreciated!
Thanks,
Jane
I tried using a linear mixed effects models in nlme, lmer, and repeated measures anovas. Every time, I run into some variation of the 'model is singular' error.
