R help - Apr 2003 - ace with mon=0

I was wondering whether someone can help me understand the
following behavior of the ace-function:

When ace is called with mon-parameter set to zero, R gives the
message "response spec can only be lin or ordered
(default)" and returns immediately. However, according to the help
file mon=0 requires the response transformations to be limited
to monotonic transformations.

For instance, the following call of the ace-function will not be
evaluated if one gives mon=0

     TWOPI <- 8*atan(1)
     x <- runif(200,0,TWOPI)
     y <- exp(sin(x)+rnorm(200)/2)
     a <- ace(x,y, mon=0)

Why does mon=0 not work and how should one interpret the message
"the response spec can only be lin or ordered" ?

Any help will be greatly appreciated,
  Christof

Christof Schuster
University of Notre Dame
Department of Psychology                       
103 Haggar Hall
Notre Dame, IN 46556

Tel: (574) 631-5473    email: cschuste at nd.edu
Fax: (574) 631-8883    www.nd.edu/~cschuste

-------------------
> 
> I was wondering whether someone can help me understand the
> following behavior of the ace-function:
> 
> When ace is called with mon-parameter set to zero, R gives the
> message "response spec can only be lin or ordered
> (default)" and returns immediately. However, according to the help
> file mon=0 requires the response transformations to be limited
> to monotonic transformations.
> 
> For instance, the following call of the ace-function will not be
> evaluated if one gives mon=0
> 
>      TWOPI <- 8*atan(1)
>      x <- runif(200,0,TWOPI)
>      y <- exp(sin(x)+rnorm(200)/2)
>      a <- ace(x,y, mon=0)
> 
> Why does mon=0 not work and how should one interpret the message
> "the response spec can only be lin or ordered" ?
> 
> Any help will be greatly appreciated,
>   Christof
> 
> Christof Schuster
> University of Notre Dame
> Department of Psychology                       
> 103 Haggar Hall
> Notre Dame, IN 46556
I use ace and avas quite a bit and have found a few tiny bugs such as 
that one.  I reported these bugs to the R bug list about a year ago 
and await corrections.  As you said, ace does handle nonmonotonic 
transformations of Y; you just have to turn off that error message.

Frank Harrell