error in parboot for occuRN

[Jeffrey Royle]

hi all,

 I have this error (below) when I try to run parboot for an occuRN model. This is strange because it seems to have been introduced in the 0.12-2 version of unmarked  (doesn't exist in 0.11 version).  

 I will fix this of course but I wonder if anyone else has encountered this error (it seems vaguely familiar to me...)

thanks

andy

 
> umfRN_b200_rev <- unmarkedFrameOccu(y=y, siteCovs=site_covs_rev)
>
> test2 <- occuRN(formula= ~ 1 ~  APP+GP + Fnat+Silvicultura+ Cana  , data= umfRN_b200_rev)
>
> chisq2 <- function(fm) {
+   observed <- getY(umfRN_b200_rev)
+   expected <- fitted(fm)
+   sum((observed - expected)^2/expected, na.rm=T)  
+ }
>
> pb_rev <- parboot(test2, statistic = chisq2, nsim = 20)
Error in .local(object, ...) :
  no slot of name "K" for this object of class "unmarkedFitOccuRN"
>

[Tyler Petroelje]

Hi Andy,

Did you ever figure out what caused this? I seem to be having a similar issue currently. Ran on both version 0.12-3 and 1.0.0

> fmRNnull <- occuRN(~1 ~1,data=c.umf,K=40,method="BFGS",se=TRUE)

>

> fitstats <- function(Mod_global2) {

+   observed <- getY(Mod_global2@data)

+   expected <- fitted(Mod_global2)

+   resids <- residuals(Mod_global2)

+   sse <- sum(resids^2,na.rm=TRUE)

+   chisq <- sum((observed - expected)^2 / expected,na.rm=TRUE)

+   freeTuke <- sum((sqrt(observed) - sqrt(expected))^2,na.rm=TRUE)

+   out <- c(SSE=sse, Chisq=chisq, freemanTukey=freeTuke)

+   return(out)

+ }

> pb <- parboot(fmRNnull, fitstats, nsim=10000, report=1)

Error in .local(object, ...) : 

  no slot of name "K" for this object of class "unmarkedFitOccuRN"

In addition: Warning messages:

1: In truncateToBinary(object@data@y) :

  Some observations were > 1.  These were truncated to 1.

2: In truncateToBinary(y) :

Thanks!

- Tyler

[Ken Kellner]

Hi Tyler,

You have to supply a K argument to the fitted function for occuRN models at the moment in order for it to work. If you don't it instead looks for a nonexistent K slot which is something we should fix. So something like the following should work:

> fitstats <- function(Mod_global2) {

+   observed <- getY(Mod_global2@data)

+   expected <- fitted(Mod_global2, K=40) #### add K here

+   resids <- residuals(Mod_global2)

+   sse <- sum(resids^2,na.rm=TRUE)

+   chisq <- sum((observed - expected)^2 / expected,na.rm=TRUE)

+   freeTuke <- sum((sqrt(observed) - sqrt(expected))^2,na.rm=TRUE)

+   out <- c(SSE=sse, Chisq=chisq, freemanTukey=freeTuke)

+   return(out)

+ }

> pb <- parboot(fmRNnull, fitstats, nsim=10000, report=1)

Ken

[Jeffrey Royle]

hi all,

 Ken is indeed right -- that will work.  I thought for sure we had a trap in the fitted() to default the value of K for that function.. At any rate, we will fix this.

regards

andy

--
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/7f79ba5f-dad4-44d7-b28e-4b8e200932e7%40googlegroups.com.

[Ken Kellner]

Turns out it isn't quite that simple, since residuals() looks for K also. I attached a script that overwrites the fitted and residual methods, if you source it in after loading unmarked your original code should work, as a workaround until we fix it.

Ken

To unsubscribe from this group and stop receiving emails from it, send an email to unma...@googlegroups.com.

[Jeffrey Royle]

interestingly we do have the fitted() function for occuRN looking for object@K from the occuRN function. The problem is, occuRN doesn't return K in the fit list!  So if we change that, then at least "fitted" will work... let me do that.  I'll also check residuals()

To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/faf415cc-d7f3-460b-93c0-92682142d075%40googlegroups.com.

[Jeffrey Royle]

I just pushed a change to occuRN to return K in the fit list... I think if you install from github then things should work out.  I will test it this weekend.

regards

andy

[Tyler Petroelje]

Many thanks, Andy and Ken,

The workaround to provide K to fitted and residuals worked. I'll pull the latest commit from github and see how that works as well.

Thanks again,

- Tyler

To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/CAE%2BwgF1a%3DhyN2_W%2BF6v%2BaRQWSbBJtFJQvsudZ6tS5Dgh316iJA%40mail.gmail.com.