reduce.capthist {secr}

Combine Occasions Or Detectors

Description

Use these methods to combine data from multiple occasions or multiple detectors in a capthist or traps object, creating a new data set and possibly converting between detector types.

Usage

## S3 method for class 'traps'
reduce(object, newtraps = NULL, newoccasions = NULL, span = NULL, 
    rename = FALSE, newxy = c('mean', 'first'), ...)

## S3 method for class 'capthist'
reduce(object, newtraps = NULL, span = NULL, rename =
    FALSE, newoccasions = NULL, by = 1, outputdetector = NULL,
    select = c("last","first","random"), dropunused = TRUE, verify
    = TRUE, sessions = NULL, ...)

Arguments

object

traps or capthist object

newtraps

list in which each component is a vector of subscripts for detectors to be pooled

newoccasions

list in which each component is a vector of subscripts for occasions to be pooled

span

numeric maximum span in metres of new detector

rename

logical; if TRUE the new detectors will be numbered from 1, otherwise a name will be constructed from the old detector names

newxy

character; coordinates when detectors grouped with ‘newtraps’

by

number of old occasions in each new occasion

outputdetector

character value giving detector type for output (defaults to input)

select

character value for method to resolve conflicts

dropunused

logical, if TRUE any never-used detectors are dropped

verify

logical, if TRUE the verify function is applied to the output

sessions

vector of session indices or names (optional)

...

other arguments passed by reduce.capthist to reduce.traps, or by reduce.traps to hclust

Details

reduce.traps –

Grouping may be specified explicitly via newtraps, or implicitly by span.

If span is specified a clustering of detector sites will be performed with hclust and detectors will be assigned to groups with cutree. The default algorithm in hclust is complete linkage, which tends to yield compact, circular clusters; each will have diameter less than or equal to span.

newxy = 'first' selects the coordinates of the first detector in a group defined by ‘newtraps’, rather then the average of all detectors in group.

reduce.capthist –

The first component of newoccasions defines the columns of object for new occasion 1, the second for new occasion 2, etc. If newoccasions is NULL then all occasions are output. Subscripts in a component of newoccasions that do not match an occasion in the input are ignored. When the output detector is one of the trap types (‘single’, ‘multi’), reducing capture occasions can result in locational ambiguity for individuals caught on more than one occasion, and for single-catch traps there may also be conflicts between individuals at the same trap. The method for resolving conflicts among ‘multi’ detectors is determined by select which should be one of ‘first’, ‘last’ or ‘random’. With ‘single’ detectors select is ignored and the method is: first, randomly select* one trap per animal per day; second, randomly select* one animal per trap per day; third, when collapsing multiple days use the first capture, if any, in each trap.

Usage data in the traps attribute are also pooled if present; usage is summed over contributing occasions and detectors. If there is no 'usage' attribute in the input, and outputdetector is one of 'count', 'polygon', 'transect' and 'telemetry', a homogeneous (all-1's) 'usage' attribute is first generated for the input.

* i.e., in the case of a single capture, use that capture; in the case of multiple ‘competing’ captures draw one at random.

If newoccasions is not provided then old occasions are grouped into new occasions as indicated by the by argument. For example, if there are 15 old occasions and by = 5 then new occasions will be formed from occasions 1:5, 6:10, and 11:15. A warning is given when the number of old occasions is not a multiple of by as then the final new occasion will comprise fewer old occasions.

dropunused = TRUE has the possibly unintended effect of dropping whole occasions on which there were no detections.

A special use of the by argument is to combine all occasions into one for each session in a multi-session dataset. This is done by setting by = "all".

reduce.capthist may be used with non-spatial capthist objects (NULL 'traps' attribute) by setting verify = FALSE.

Value

reduce.traps –

An object of class traps with detectors combined according to newtraps or span. The new object has an attribute ‘newtrap’, a vector of length equal to the original number of detectors. Each element in newtrap is the index of the new detector to which the old detector was assigned (see Examples).

The object has no clusterID or clustertrap attribute.

reduce.capthist –

An object of class capthist with number of occasions (columns) equal to length(newoccasions); detectors may simulataneously be aggregated as with reduce.traps. The detector type is inherited from object unless a new type is specified with the argument outputdetector.

Warning

The argument named ‘columns’ was renamed to ‘newoccasions’ in version 2.5.0, and arguments were added to reduce.capthist for the pooling of detectors. Old code should work as before if all arguments are named and ‘columns’ is changed.

Note

The reduce method may be used to re-assign the detector type (and hence data format) of a capthist object without combining occasions or detectors. Set the object and outputdetector arguments and leave others at their default values.

Automated clustering can produce unexpected outcomes. In particular, there is no guarantee that clusters will be equal in size. You should inspect the results of reduce.traps especially when using span.

reduce.traps is not implemented for polygons or transects.

The function discretize converts polygon data to point-detector (multi, proximity or count) data.

See Also

capthist, subset.capthist, discretize, hclust, cutree

Examples

tempcapt <- sim.capthist (make.grid(nx = 6, ny = 6), nocc = 6)
class(tempcapt)

pooled.tempcapt <- reduce(tempcapt, newocc = list(1,2:3,4:6))
summary (pooled.tempcapt)

pooled.tempcapt2 <- reduce(tempcapt, by = 2)
summary (pooled.tempcapt2)

## collapse multi-session dataset to single-session 'open population'
onesess <- join(reduce(ovenCH, by = "all"))
summary(onesess)

# group detectors within 60 metres
plot (traps(captdata))
plot (reduce(captdata, span = 60), add = TRUE)

# plot linking old and new
old <- traps(captdata)
new <- reduce(old, span = 60)
newtrap <- attr(new, "newtrap")
plot(old, border = 10)
plot(new, add = TRUE, detpar = list(pch = 16), label = TRUE)
segments (new$x[newtrap], new$y[newtrap], old$x, old$y)

## Not run: 

# compare binary proximity with collapsed binomial count
# expect TRUE for each year
for (y in 1:5) {
    CHA <- abs(ovenCHp[[y]])   ## abs() to ignore one death
    usage(traps(CHA)) <- matrix(1, 44, ncol(CHA))
    CHB <- reduce(CHA, by = 'all', output = 'count')
    # summary(CHA, terse = TRUE)
    # summary(CHB, terse = TRUE)
    fitA <- secr.fit(CHA, buffer = 300, trace = FALSE)
    fitB <- secr.fit(CHB, buffer = 300, trace = FALSE, binomN = 1, biasLimit = NA)
    A <- predict(fitA)[,-1] 
    B <- predict(fitB)[,-1]
    cat(y, ' ', all(abs(A-B)/A < 1e-5), '\n')
}
## multi-session fit
## expect TRUE overall
CHa <- ovenCHp
for (y in 1:5) {
    usage(traps(CHa[[y]])) <- matrix(1, 44, ncol(CHa[[y]]))
    CHa[[y]][,,] <- abs(CHa[[y]][,,])
}
CHb <- reduce(CHa, by = 'all', output = 'count')
summary(CHa, terse = TRUE)
summary(CHb, terse = TRUE)
fita <- secr.fit(CHa, buffer = 300, trace = FALSE)
fitb <- secr.fit(CHb, buffer = 300, trace = FALSE, binomN = 1, biasLimit = NA)
A <- predict(fita)[[1]][,-1] 
B <- predict(fitb)[[1]][,-1]
all(abs(A-B)/A < 1e-5)


## End(Not run)