[R] problem with fitdistr ?

[Peter Dalgaard]

Prof Brian Ripley <ripley at stats.ox.ac.uk> writes:

> Which version of MASS (not that you gave me any credit)?  This works in 
> the current 7.1-14.

Odd things are happening for me with r-devel, though:

> library(MASS)
> fitdistr(rnorm(100),"Normal")
Error in dn[[2]] : subscript out of bounds
> fitdistr(rnorm(100),"Normal",start=list(mean=0,sd=1))
Error in fitdistr(rnorm(100), "Normal", start = list(mean = 0, sd =
> 1)) :
        supplying pars for the Normal is not supported
> x <- fitdistr(rnorm(100),"normal")
> x
Error in dn[[2]] : subscript out of bounds
> str(x)
List of 2
 $ estimate: num 0.217
 $ sd      : num 0.99
 - attr(*, "class")= chr "fitdistr"

And the rest of the story is that this bit of print.fitdistr computes
"ans" without dimnames and thus refers to dn[[2]] before there's
anything there:

    ans <- format(rbind(x$estimate, x$sd), digits = digits)
    ans[1, ] <- sapply(ans[1, ], function(x) paste("", x))
    ans[2, ] <- sapply(ans[2, ], function(x) paste("(", x, ")",
        sep = ""))
    dn <- dimnames(ans)
    dn[[1]] <- rep("", 2)
    dn[[2]] <- paste(substring("      ", 1, (nchar(ans[2, ]) -
        nchar(dn[[2]]))%/%2), dn[[2]])


 

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907