I'm trying to fit a mixed model to see whether or not the change in my measurements (variable "diff") is significantly different from zero (taking to account subject's individual effects). Here's my data frame:
> dput(mixed)
structure(list(subject.id = c(4, 4, 4, 4, 6, 6, 6, 6, 9, 9, 9,
9, 12, 12, 12, 12, 13, 13, 13, 13, 15, 15, 15, 15, 19, 19, 19,
19, 21, 21, 21, 21, 23, 23, 23, 23, 28, 28, 28, 28), diff = c(-1,
-3, -5, 4, 0, -1, 1, -11, 0, -9, 11, -2, 0, -1, -4, 0, -3, -2,
-1, -4, -8, 2, -5, NA, 3, -2, -3, -3, -5, -9, 3, -2, 1, 2, 9,
-17, 25, -9, NA, NA)), row.names = c(NA, -40L), class =
"data.frame")
subject.id diff
1 4 -1
2 4 -3
3 4 -5
4 4 4
5 6 0
6 6 -1
7 6 1
8 6 -11
9 9 0
10 9 -9
11 9 11
12 9 -2
13 12 0
14 12 -1
15 12 -4
16 12 0
17 13 -3
18 13 -2
19 13 -1
20 13 -4
21 15 -8
22 15 2
23 15 -5
24 15 NA
25 19 3
26 19 -2
27 19 -3
28 19 -3
29 21 -5
30 21 -9
31 21 3
32 21 -2
33 23 1
34 23 2
35 23 9
36 23 -17
37 28 25
38 28 -9
39 28 NA
40 28 NA
Basically, I have 4 days of pre and post measurements for my 10 subjects. The above data comes from the difference between pre and post measurements. Here's my code:
mod.crossshift <- lmer(formula =diff ~ 1 + (1|subject.id), data =
mixed, REML = FALSE)
When I run the model, I do see the results, but I also get an error: "boundary (singular) fit: see ?isSingular"
Upon some google search, it seems like this happens sometimes because the model is overcomplicated, but as far as I know, this one is as simple as a model can get.
What does this mean and can I ignore/fix it?
