informR (version 1.0-5)

gen.sformlist: Generate a list of S-form sequence statistics.

Description

Generates a list of S-form sequence statistics based on sform regular expressions. Output is not used directly but is passed to other methods.

Usage

gen.sformlist(evl, sforms,cond=FALSE, interval=FALSE, warn=TRUE,...)

Arguments

evl

An eventlist passed from gen.evl()

sforms

A character string of S-form regular expressions representing the sequence statistics.

cond

Should the resulting statistics condition out the prefix? Defaults to FALSE.

interval

When cond=TRUE, are elements of sforms paired as start-stop terms?. Defaults to FALSE.

warn

Should warnings be issued? Defaults to TRUE.

...

Additional arguments to pass to gen.sform

Value

• A list containing idXevent, iXj matrices of sform sequence statistics.

Details

Regular expressions must adhere to PERL standards and, at this time, must make use of the alphabetic evl$event.key values in an eventlist object. Two special regex operators are permitted in S-form expressions: the ``|'' (OR operator) and the ``+'' (AND operator). In S-form expressions, the OR operator is used to differentiate between two possible event paths and the AND operator is used to indicate persistence of the last event.

Useful expressions include: rll{ [,1] aa inertial term: S-form of the type ``event a predicts event a'' [,2] ab basic digram transition term: S-form of the type ``event a predicts event b'' [,3] a+b transition term with persistence: S-form of the type ``some series of events a predicts event b'' [,4] abc basic trigram term: S-form of the of the type ``event a followed by event b predicts event c'' [,5] aab tuples term: S-form of the of the type ``event a followed by event a predicts event c'' [,6] (a|b)c complex term with divergence: S-form of the of the type ``event a OR event b predicts event c'' [,7] (a|b+d)c complex term with divergence and persistence: S-form of the of the type ``event a predicts event c OR some series of events b followed by event d predicts event c'' }

There may be more than one way to form an S-form expression. For example, aa is equivalent to a+a; however, the former is preferred because the search methods used to update the sufficient statistics matrix are much faster in that case. This is also important to note because the informR package does not check for affine collinearity between the sufficient statistics (i.e., as would be the case in gen.sformlist(evl, c("aa","a+a","aaa")).

To generate S-form statistics that condition out the prefix set cond=TRUE. This will result in statistics of the form "(aa)b" or the likelihood of "b" given that "aa" occurred; thus, only the hazard of "b" is affected. This is useful when only the suffix event is of interest. When elements of each S-form are "paired," such as interval likelihood cases where each spell class can be characterized as having ``starting'' and ``stopping'' elements, setting interval=TRUE will result in two statistics per s-form: one that models the hazard of starting the suffix spell and one that models the duration of the suffix spell. In the general case, this is equivalent to a model with terms for "(ab)c" and "(ab)cd", respectively.

See Also

gen.sform,sfl2sl,glb.sformlist

Examples

set.seed(57391)
rawevents<-sample(rep(c("ran","eat","stay","eat","ran","play"),50))
actors<-rep(c("Jim","Bill","Pete"),100)
evmat<-cbind(rawevents,actors)
eventlist<-gen.evl(evmat)

#See event-type alphabetic token codes
eventlist$event.key

alpha.ints<-gen.intercepts(eventlist,basecat="ran")

#Create some inertial s-form statistics and fit the models
beta.sforms<-gen.sformlist(eventlist,c("aa","bb","cc","dd"))

#Condition out the effects of the prefix:
gamma.sforms<-gen.sformlist(eventlist,c("aab","abb","acc","add"),cond=TRUE)

#Manual example of the above
sforms1<-c("aab","abb","acc","add")
sforms2<-sapply(sforms1,function(x) substr(x,1,nchar(x)-1))

sforms1.sf<-gen.sformlist(eventlist, sforms1)
sforms2.sf<-gen.sformlist(eventlist, sforms2)

for(i in 1:length(sforms1.sf)){
   for(j in 1: dim(sforms1.sf[[1]])[[3]]){
      sforms1.sf[[i]][,,j]<-abs(sforms1.sf[[i]][,,j]-sforms2.sf[[i]][,,j])
   }
}

#Note the difference:
gamma.sforms2<-gen.sformlist(eventlist,c("aab","abb","acc","add"))
gamma.sforms2[[1]][75:85,,1]
gamma.sforms[[1]][75:85,,1]