[R-sig-eco] plyr and mvabund, conceptual issue

[Hadley Wickham]

On Wed, Oct 29, 2014 at 4:23 AM, Eduard Szöcs <eduardszoecs at gmail.com> wrote:
> Hai Kendra,
>
> i've used a simple for-loop to do this in the past.
>
> Something along these lines:
>
>
> ###-----------------------------------------------------------------
> mymv <- function(response, env, zone) {
>   df <- data.frame(env = env, zone = zone)
>   out <- NULL
>   for (i in levels(zone)) {
>     rsp <- mvabund(response[zone == i, ])
>     out[[i]]$mod <- manyglm(rsp ~ env, data = df[zone == i, ])
>     out[[i]]$anova <- anova(out[[i]]$mod, p.uni = "adjusted", resamp =
> "perm.resid", nBoot = 10)
>   }
>   return(out)
> }

If you're going to use a for-loop, you need to preallocate the output:

out <- vector("list", length(levels(zone))

Otherwise each iteration of the loop needs to copy all previous output
to a new location, making things rather slow.

Hadley

-- 
http://had.co.nz/