R help - May 2008 - Data manipulation for random intercept GLMM

Hello,

I am working on some examples of GLMM for my students but I am afraid
that my way of preparing a dataframe to pass to lmer will make them
think that R is a very difficult and un-natural language. Here is for
example a simple data set about approval ratings on two different
surveys for a random sample of 1600 individuals. 
> ## Example: Ratings of prime minister (Agresti, Table 12.1, p.494)
> rating <- matrix(c(794, 86, 150, 570), 2, 2)
> dimnames(rating) <- list(First = c("approve",
"disapprove"),
+                          Second = c("approve",
"disapprove"))
> rating
            Second
First        approve disapprove
  approve        794        150
  disapprove      86        570

It seems to me that, in order to fit a model using lmer, I cannot use
the table directly, but I need a dataframe with 1600 x 2 rows and
columns response (Approve/Disapprove), survey (First/Second), and
subject id. So I proceeded to create such a dataframe:
> approval <- factor(c("Approve", "Disapprove"),
+                    levels = c("Disapprove",
"Approve"))
> survey <- factor(c("First", "Second"))
> tmp <- data.frame(approval = unlist(expand.grid(approval, approval)),
+                   survey = rep(survey, each = 4))
> rat.df <- cbind(tmp[rep(1:8, rep(rating, 2)), ],
+                 id = factor(rep(1:sum(rating), 2)))
> row.names(rat.df) <- NULL
That does the job, since now I can call lmer:
> m1 <- lmer(approval ~ survey + (1 | id), family = binomial, data =
rat.df,
+            method = "Laplace")

The issue I have is that creating the 'rat.df' dataframe above will
likely make all of my students look for a different software. So my
question is the following. Is there a more elegant way to create the
dataframe needed by lmer from the tabular form in which one is more
likely to find these kind of data?

Consider also that the next simplest example is the following, in
which there are three items on a questionnaire and gender is included
in the model:
> ### Example: Support for legalizing abortion (Agresti, Table 10.13, p.441)
> legalize <- matrix(c(342, 440, 26, 25, 6, 14, 21, 18, 11, 14,
+                      32, 47, 19, 22, 356, 457), nr =
2)
> dimnames(legalize) <- list(Gender = c("Male",
"Female"),
+                            Three = c(111, 112, 211, 212, 121,
+                            122, 221, 222))
> legalize
        Three
Gender   111 112 211 212 121 122 221 222
  Male   342  26   6  21  11  32  19 356
  Female 440  25  14  18  14  47  22 457

(Here '111' means (Yes, Yes, Yes) on the three items, etc.)

How can I tranform elegantly this table into a dataframe that I can
feed to lmer?

Thank you in advance for your replies!

Giovanni Petris

-- 

Giovanni Petris  <GPetris at uark.edu>
Department of Mathematical Sciences
University of Arkansas - Fayetteville, AR 72701
Ph: (479) 575-6324, 575-8630 (fax)
http://definetti.uark.edu/~gpetris/

Giovanni Petris <GPetris <at> uark.edu> writes:
> > ## Example: Ratings of prime minister (Agresti, Table 12.1, p.494)
> > rating <- matrix(c(794, 86, 150, 570), 2, 2)
> > dimnames(rating) <- list(First = c("approve",
"disapprove"),
> +                          Second = c("approve",
"disapprove"))
> > rating
>             Second
> First        approve disapprove
>   approve        794        150
>   disapprove      86        570
>
[snip]
> 
> > approval <- factor(c("Approve", "Disapprove"),
> +                    levels = c("Disapprove",
"Approve"))
> > survey <- factor(c("First", "Second"))
> > tmp <- data.frame(approval = unlist(expand.grid(approval,
approval)),
> +                   survey = rep(survey, each = 4))
> > rat.df <- cbind(tmp[rep(1:8, rep(rating, 2)), ],
> +                 id = factor(rep(1:sum(rating), 2)))
> > row.names(rat.df) <- NULL
> 
> That does the job, since now I can call lmer:
> 
> > m1 <- lmer(approval ~ survey + (1 | id), family = binomial, data =
rat.df,
> +            method = "Laplace")
> 
[snip]
> 
> Consider also that the next simplest example is the following, in
> which there are three items on a questionnaire and gender is included
> in the model:
> 
> > ### Example: Support for legalizing abortion (Agresti, Table 10.13,
p.441)
> > legalize <- matrix(c(342, 440, 26, 25, 6, 14, 21, 18, 11, 14,
> +                      32, 47, 19, 22, 356, 457), nr = 2)
> > dimnames(legalize) <- list(Gender = c("Male",
"Female"),
> +                            Three = c(111, 112, 211, 212, 121,
> +                            122, 221, 222))
> > legalize
>         Three
> Gender   111 112 211 212 121 122 221 222
>   Male   342  26   6  21  11  32  19 356
>   Female 440  25  14  18  14  47  22 457
> 
> (Here '111' means (Yes, Yes, Yes) on the three items, etc.)
> 
> How can I tranform elegantly this table into a dataframe that I can
> feed to lmer?
> 
  I would like to know the answers to this as well.  I thought
I was going to be able to get away with as.data.frame.table()
[which is handy and underappreciated], but it doesn't expand
counts.
   I hope you get an elegant solution: in the meantime, I guess
my advice is to try to write a reasonably general (if not necessarily
absolutely elegant) function that you can supply your students with.
If they have an expand.table() function, they won't care how complex
it is internally ...

  Ben Bolker