Predict function with glmer

[abasbabatunde]

I am trying to use the predict function to estimate a dependent variable in my mixed model given value of the two independent variables that are in the model. The estimable function in lmer for some reasons doesn't work with glmer. Can someone help with how I could estimate this? see my code. I got error that my object is not a matrix.

model1=glmer(response~treatment+month+treatment*month +
(1|id),
data=data1,
family=binomial(link = "logit"),
nAGQ=100)
summary(model1)

nedff = data.frame(treatment =1, month =12)
predict(model1, nedff, se.fit=T, interval ="confidence")

[gueyenono]

Hello @abasbabatunde ,

It would be much easier to help you if your example was reproducible. The fact that we don't have your data (or a piece of it) makes it hard to debug your code on your end. If you want to know more about asking a reproducible question, please check this:

FAQ: What's a reproducible example (`reprex`) and how do I create one?

Why reprex? Getting unstuck is hard. Your first step here is usually to create a reprex, or reproducible example. The goal of a reprex is to package your code, and information about your problem so that others can run it and feel your pain. Then, hopefully, folks can more easily provide a solution. What's in a Reproducible Example? Parts of a reproducible example: background information - Describe what you are trying to do. What have you already done? complete set up - include any library() calls and data to reproduce your issue. data for a reprex: Here's a discussion on setting up data for a reprex make it run - include the minimal code required to reproduce your error on the data…

[abasbabatunde]

Poor me! I went through the link you shared and I still didn't know how to use the reprex. I am finding it difficult to get my code chunk as it is and post here or my data. Any other means to help?

[gueyenono]

You could copy and paste the output of:

dput(data1)

Then we should be able to use it to recreate your dataset. If the data is too big, then you could only share the first 20 lines (for exemple). The code for that would be:

dput(head(data1, 20))

[abasbabatunde]

Thank you so much. I will reach out later to know how to do reprex because R is going to be my main language soon. here is the output from dput(head(data1,20).

dput(head(data1,20))
structure(list(V1 = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L,
2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L), V2 = c(1L, 1L, 1L, 0L,
0L, 0L, 0L, 0L, 0L, 1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L
), V3 = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L), V4 = c(0, 0.8571428571, 3.5357142857,
4.5357142857, 7.5357142857, 10.035714286, 13.071428571, 0, 0.9642857143,
2, 3.0357142857, 6.5, 9, 0, 1.25, 2.1071428571, 3.25, 6.75, 9.0357142857,
11.5), V5 = c(1L, 2L, 3L, 4L, 5L, 6L, 7L, 1L, 2L, 3L, 4L, 5L,
6L, 1L, 2L, 3L, 4L, 5L, 6L, 7L)), row.names = c(NA, 20L), class = "data.frame")

Here is my glmer code
model1=glmer(response~treatment+month+treatment*month +
(1|id),
data=data1,
family=binomial(link = "logit"),
nAGQ=100)
summary(model1)

I am trying to estimate response when treatment is 1 and month is 12. This is not working.

nedff = data.frame(treatment =1, month =12)
predict(model1, nedff, se.fit=T, interval ="confidence")

[nirgrahamuk]

the glmer code you wrote with respect to data1 could not work with the data1 example you gave us; what you gave us does not have the named columns response, treatment etc. it has V1,V2,V3, etc..
please review

[abasbabatunde]

Apoligies, here is the clean data

dput(head(data1,20))
structure(list(id = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L,
2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L), response = c(1L, 1L,
1L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 1L), treatment = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), month = c(0, 0.8571428571,
3.5357142857, 4.5357142857, 7.5357142857, 10.035714286, 13.071428571,
0, 0.9642857143, 2, 3.0357142857, 6.5, 9, 0, 1.25, 2.1071428571,
3.25, 6.75, 9.0357142857, 11.5), visit = c(1L, 2L, 3L, 4L, 5L,
6L, 7L, 1L, 2L, 3L, 4L, 5L, 6L, 1L, 2L, 3L, 4L, 5L, 6L, 7L)), row.names = c(NA,
20L), class = "data.frame")

[nirgrahamuk]

@abasbabatunde

predict(model1, nedff, se.fit=T, interval ="confidence")

so, so the error about your data not being a matrix is unfortunately unclear; but its caused by your model formula involving id; which means predict wants to see an id column

[abasbabatunde]

That is unfortunate because the ID is needed as the random effect in the model.
Which other way do you think I can predict or get estimates based on the values of the predictors?

[nirgrahamuk]

I dont think theres harm simply adding an id field to your new data. But maybe someone more knowledgeable than me will chime in.