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

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

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

Solution

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}