Skip to content
Prev 17901 / 20628 Next

Understanding and analysing split split design using lmer()

Dear Peter,

Your reply has made my day and very clearly answered the questions I was
asking myself for a few weeks. Thank you for correcting my mistakes as
well.

The lmer() does produce a variance of 0 for random effect (crop: rep). I
now know the term is called boundary (singular) fit. What would be the
correct thing to do in this case? I do have this experiment conducted for 2
years in 2 different locations, which might take care of the precious
degrees of freedom, but again, will add new variables making the model more
complicated.

I ran an Anova II (wald test) and was thinking of doing post hoc analysis
on significant terms. (If this would take care of singularity). Any advice
on that would be greatly appreciated.

And yes, I am in Agronomy in U of M. I will be writing to you in your work
email about the friendly statistician you mentioned!!!!


Best regards,
Rabin
On Mon, Sep 9, 2019 at 4:43 PM Peter Claussen <dakotajudo at mac.com> wrote: