Message-ID: <22849.25327.306098.178721@losangelesyouthorchestra.org>
Date: 2017-06-14T16:23:11Z
From: Don Cohen
Subject: understanding I() in lmer formula
In-Reply-To: <CAET4i1dJao=jj4Kj+C9sRRV4w2XOMS7f2fbJX5ZvOfEXBJYMvA@mail.gmail.com>
Dan Brooks writes:
> Related to your second question about ||, I believe || is shorthand for
> uncorrelating the slope and intercept, e.g:
> fm3 <- lmer(Reaction~Days+(Days+0|Subject)+(1|Subject),data=sleepstudy)
> fm4 <- lmer(Reaction~Days+(1+Days||Subject),data=sleepstudy)
>
> As opposed to:
> fm2 <- lmer(Reaction~Days+(Days+1|Subject),data=sleepstudy)
> or simply:
> fm1 <- lmer(Reaction~Days+(Days|Subject),data=sleepstudy)
Saying that it's shorthand for something else would be useful if I
understood what that something else meant!
I gather the formula language is intended to be intuitive, and maybe
it is for people who understand mixed models better than I do, but
to me fm3 and fm4 seem equally mysterious. The idea that days+1 should
be the same as days and that days+0 should be different from days seems
counter-intuitive to me.
And the distinction between correlated vs uncorrelated also still
has to be explained - I described some of my interpretation
problems in the previous post.