Problem with effects function : how to get adjusted value from a model?

Question

I encounter using "effect" from effects package. My goal is to extract the values of the fitted model for some values of my predictor. The problem is for a mixed model with quadratic predictors but it can be reproduced with a simple linear model.

Let's consider the following generated data :

y=rnorm(n = 12,mean = 10,sd = 3)
time=rep(c(0,1,6,10),3)
df=data.frame(time,y)

Now I fit a linear model and get the effect of "time" with the effect function :

model=lm(y~time, data=df)
fx=effect("time", model) 

I was assuming I could get the values of the model for my 4 time values with :

time=fx$x
values=fx$fit

But instead of my 4 time points fx$x returns a time vector with : 0 2 5 8 10

I don't understand why (but my guess is that I'm doing it wrong). I just wand to get the values of the model for my time points (0,1,6,10). How can I do (that will work too with a lme4::lmer model)?

Thanks in advance

Answer 1

By default, the effects package spaces values evenly over the range of the focal predictor. You can specify the values you want explicitly in the xlevels argument:

library(effects)
#> Loading required package: carData
#> lattice theme set by effectsTheme()
#> See ?effectsTheme for details.
y=rnorm(n = 12,mean = 10,sd = 3)
time=rep(c(0,1,6,10),3)
df=data.frame(time,y)


model=lm(y~time, data=df)
fx=effect("time", model, 
          xlevels = list(time = unique(df$time)) )
time=fx$x
time
#>   time
#> 1    0
#> 2    1
#> 3    6
#> 4   10
values=fx$fit
values
#>       [,1]
#> 1 9.629158
#> 2 9.439073
#> 3 8.488650
#> 4 7.728311

^{Created on 2024-01-26 with reprex v2.0.2}