logLik {PtProcess} | R Documentation |
Calculates the log-likelihood of a point process. Provides methods for the generic function logLik
.
## S3 method for class 'mpp': logLik(object, SNOWcluster=NULL, ...) ## S3 method for class 'linksrm': logLik(object, ...)
object |
an object with class "mpp" or "linksrm" . |
SNOWcluster |
an object of class "cluster" created by the package snow; default is NULL . Enables parallel processing if not NULL . See “Parallel Processing” below for further details. |
... |
other arguments. |
Value of the log-likelihood.
Parallel processing can be enabled to calculate the term sum_i log lambda_g(ti|H_ti). Generally, the amount of computational work involved in calculating lambda_g(t|H_t) is much greater if there are more events in the process history prior to t than in the case where there are fewer events. Given m nodes, the required evaluation points are divided into m groups, not of equal number, but of roughly equal amount of work, taking into account the amount of “history” prior to each event. Currently this division is very basic, and assumes that all nodes run at a roughly comparable speed.
If the communication between nodes is slow and the dataset is small, then the time taken to allocate the work to the various nodes may in fact take more time than simply using one processor to perform all of the calculations.
The required steps in initiating parallel processing are as follows.
# load the "snow" package library(snow) # define the SNOW cluster object. It contains names and # possibly other information about the available nodes cl <- makeSOCKcluster(c("localhost", "localhost", "horoeka.localdomain", "horoeka.localdomain")) # then define the required model object, e.g. see topic "mpp" # say the model object is called x # then calculate the log-likelihood as print(logLik(x, SNOWcluster=cl)) # stop the R jobs on the slave machines stopCluster(cl)Note that the communication method does not need to be
SOCKS
; see the snow package documentation, topic snow-startstop
, for other options. Further, if some nodes are on other machines, the firewalls may need to be tweaked. The master machine initiates the R jobs on the slave machines by communicating through port 22 (use of security keys are needed rather than passwords), and subsequent communications through port 10187. Again, these details can be tweaked in the options settings within the snow package.
# SRM: magnitude iid exponential with bvalue=1 TT <- c(0, 1000) bvalue <- 1 params <- c(-2.5, 0.01, 0.8, bvalue*log(10)) # calculate log-likelihood excluding the mark density term x1 <- mpp(data=NULL, gif=srm_gif, mark=list(NULL, rexp_mark), params=params, gmap=expression(params[1:3]), mmap=expression(params[4]), TT=TT) x1 <- simulate(x1, seed=5) print(logLik(x1)) # calculate log-likelihood including the mark density term x2 <- mpp(data=x1$data, gif=srm_gif, mark=list(dexp_mark, rexp_mark), params=params, gmap=expression(params[1:3]), mmap=expression(params[4]), TT=TT) print(logLik(x2)) # contribution from magnitude marks print(sum(dexp(x1$data$magnitude, rate=bvalue*log(10), log=TRUE)))