gap_statistic: Computing the Gap Statistic

In jmbh/mta: Processing of 2d movement data from 2-choice mousetracking experiments

Description

Computing within cluster dissimilarity for different numbers of cluster k and the Gap statistic.

Usage

gap_statistic(dist, kseq, steps, lambda = .7, bezier = NULL, 
              xcor = c(0,1,-1),  ycor = c(0,1.5,1.5))

Arguments

dist	p x p distance matrix, where p is the number of trajectories.
kseq	A sequence of cluster sized, for which the within cluster dissimilarity should be computed
steps	The number of data points on the generated trajectories.
lambda	Lambda parameter of the exponential distribution, from which the parameters for the Bezier curve are drawn, which we use to generate trajectories. To get a feel for this parameter, try out values and look at the generated data.
bezier	Provide a custom sequence of Bezier curve parameters.
xcor	x-coordinates of the start point, the non-selected box and the end-point in that order.
ycor	y-coordinates of the start point, the non-selected box and the end-point in that order.

Value

The function returns a list:

WCD_data	Within cluster dissimilarity for each k for the real data.
WCD_syn	Within cluster dissimilarity for each k for the generated data.
Gaps	The difference values of WCD_syn - WCD_data
cl_data	Cluster assignments for each k
cl_syn	Cluster assignments for each k and the generated data

Author(s)

Jonas Haslbeck <[e-mail]>

Examples

## Not run: 

# Example data
head(data_sp2015)

# Preprocess Data
stre <- list("start"=c(0,0), "left"=c(-1,1.5), "right"=c(1,1.5))
steps <- 5
data_pr <- prepr(data_sp2015, i.xyt = c('x', 'y', 't'), i.id=c('id.ptp', 'id.trial'), 
                  type='spatial', stretch = stre, steps = steps)$data
data_pr$x[data_pr$chosen.box==1] <- data_pr$x[data_pr$chosen.box==1] * -1 # all endpoints to one side

# Compute distance matrix
id <- 1:1344 # no of unique trials
dist <- mta:::distmat(id, data_pr$x , data_pr$y, steps)

# Call Gap statistic function
gapobj <- gap_statistic(dist, kseq=1:10, steps=5)




## End(Not run)

jmbh/mta documentation built on May 19, 2019, 1:51 p.m.