csvToUMF - Site-level covariates ?

[Ben Nickley]

Hi,

I am new to unmarked and R, so I am having some difficulties getting started.  Any help would be greatly appreciated.

What I would like to do is use the csvToUMF() function to get my data, which is in LONG format, into an unmarkedFrame.  However, when I create the unmarked frame with csvToUMF(), all of my covariates are placed at the observation level.  Is there a way to specify site-level covariates using the csvToUMF() function?  

Below is the code I ran with a subset of my data.  The whole data set is much larger and includes many NAs.  I would like AREA and Route to be site-level covariates, and Time, Temp, Wind and Cloud to be observation-level covariates.  

> bacs=read.csv("bacsTEST.csv")

> head(bacs)

  PointName Date TotalBACS AREA Route Time Temp Wind Cloud

1        J1   55         1    5     1  734   56    1     0

2        J1   56         0    5     1  602   52    0     0

3        J1   57         0    5     1  606   62    0    85

4       J12   55         0    5     0 1045   74    1     1

5       J12   56         0    5     0  703   55    0     0

6       J12   57         0    5     0  652   65    0    85

> test.umf <- csvToUMF(filename="bacsTEST.csv", long = TRUE, type = "unmarkedFramePCount")

> summary(test.umf)

unmarkedFrame Object

5 sites

Maximum number of observations per site: 3 

Mean number of observations per site: 3 

Sites with at least one detection: 3 

Tabulation of y observations:

   0    1 <NA> 

  11    4    0 

Observation-level covariates:    ### All my covariates are at the observation level.  I want AREA and Route at the site level###

      AREA         Route          Time             Temp      

 Min.   :2.0   Min.   :0.0   Min.   : 602.0   Min.   :52.00  

 1st Qu.:2.0   1st Qu.:0.0   1st Qu.: 704.0   1st Qu.:56.50  

 Median :5.0   Median :1.0   Median : 734.0   Median :65.00  

 Mean   :3.8   Mean   :0.6   Mean   : 794.5   Mean   :64.07  

 3rd Qu.:5.0   3rd Qu.:1.0   3rd Qu.: 932.5   3rd Qu.:69.00  

 Max.   :5.0   Max.   :1.0   Max.   :1056.0   Max.   :80.00  

      Wind          Cloud         JulianDate

 Min.   : 0.0   Min.   : 0.00   Min.   :55  

 1st Qu.: 0.0   1st Qu.: 0.00   1st Qu.:55  

 Median : 0.0   Median : 1.00   Median :56  

 Mean   : 1.2   Mean   :34.47   Mean   :56  

 3rd Qu.: 0.5   3rd Qu.:80.00   3rd Qu.:57  

 Max.   :12.0   Max.   :90.00   Max.   :57  

[Jeffrey Royle]

hi Ben,

 I would remove the sitecovs from the data file and read them in as a seperate nsites x 2 matrix, say X and then when you create the unmarkedFrame do this

 

siteCovs(test.umf)<- X

 

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.
For more options, visit https://groups.google.com/d/optout.

[Ben Nickley]

Hey, that worked!  Thanks Andy.

[Matt Giovanni]

Hi Andy. I tried adding sitecovs to a umf after creating with csvToUMF, but then R crashes every time I fit a basic occu model. Any tips? Thanks for your help as always.

data=read.csv("data.csv",header=TRUE)

summary(data)

sitecovs=data.frame(Length=data$Length,Facility=data$Facility,Month=data$Month)

summary(sitecovs)

umf=csvToUMF("data.csv", long = TRUE, type = "unmarkedFrameOccu")

siteCovs(umf)=data2

summary(umf)

mod=occu(~1 ~1,umf)

Matt

[Jeffrey Royle]

hi Matt,

 Sorry I don't have any ideas off the top of my head. I'd have to roll up my sleeves and get into it but I don't have time in the short term here to do that. If you make progress or come up with some ideas let us know please...

regards

andy

[Dan Linden]

Hi Matt,

You don't have data2 anywhere in your script, so I'm not sure what it looks like.  Also, "Month" seems to vary across observations so you wouldn't want that as a site-level covariate.  But you could grab length and facility from the obsCovs matrix created by csvToUMF and assign them as siteCovs:

siteCovs(umf) <- data.frame(Length=as.numeric(obsCovs(umf,matrices=T)$Length[,1]),
                            Facility=obsCovs(umf,matrices=T)$Facility[,1])

Now, with only 13 sites you will have problems fitting occupancy covariates.  In addition, you have only 8 detections from 1092 observations.  So this data set will undoubtedly give you problems with model fitting (null suggests psi=1 and p=0.007).