mcClustering: Markov Chain Clustering With And Without Mixtures-of-Experts...
In bayesMCClust: Mixtures-of-Experts Markov Chain Clustering and Dirichlet Multinomial Clustering
Description
Usage
Arguments
Details
Value
Prior Data
Reporting Progress (Log Protocol)
Warning
Note
Author(s)
References
See Also
Examples
Description
This function provides Markov chain clustering with or without multinomial logit model (mixtures-of-experts)
extension (see References). That is an MCMC sampler for the mixtures-of-experts extension of Markov chain
clustering. It requires four mandatory arguments: Data, Prior, Initial and Mcmc; each
representing a list of (mandatory) arguments: Data contains data information, Prior contains prior
information, Initial contains information about starting conditions (initial values) and Mcmc contains
the setup for the MCMC sampler.
Usage
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
mcClust(
Data = list(
dataFile = stop(
"'dataFile' (=> Njk.i) must be specified: either 'filename' (path) or data"),
storeDir = "try01", priorFile = NULL),
Prior = list( H = 4, e0 = 4, c = 1, cOff = 1, usePriorFile = FALSE,
xiPooled = FALSE, N0 = 5),
Initial = list( xi.start.ind = 3, pers = 0.7, S.i.start = NULL),
Mcmc = list( M = 50, M0 = 20, mOut = 5, mSave = 10, seed = 12345))
mcClustExtended(
Data = list(
dataFile = stop(
"'dataFile' (=> Njk.i) must be specified: either 'filename' (path) or data"),
storeDir = "try01", priorFile = NULL,
X = stop("X (matrix of covariates) must be specified")),
Prior = list( H = 4, c = 1, cOff = 1, usePriorFile = FALSE,
xiPooled = FALSE, N0 = 5, betaPrior = "informative",
betaPriorMean = 0, betaPriorVar = 1),
Initial = list( xi.start.ind = 3, pers = 0.7,
S.i.start = rep(1:H, N), Beta.start = NULL),
Mcmc = list( M = 50, M0 = 20, mOut = 5, mSave = 10,
seed = 12345))
Arguments
Data
a list consisting of: dataFile, storeDir, priorFile, X. See Details.
Prior
a list consisting of: H, e0, c, cOff, usePriorFile, xiPooled, N0, betaPrior, betaPriorMean,
betaPriorVar. See Details.
Initial
a list consisting of: xi.start.ind, pers, S.i.start, Beta.start. See Details.
Mcmc
a list consisting of: M, M0, mOut, mSave, seed. See Details.
Details
Note that the values of the arguments indicated here have nothing to do with default values! For a call of these
functions this lists-of-arguments structure requires a complete specification of all arguments!
The following arguments which are lists have to be completely provided (note that there are no such things as
default values within lists!):
Data contains:
dataFileA 3-dim array having the transition counts/frequencies structure (like Njk.i
in the example data sets) already loaded into the current environment/workspace. Or a character with the name of or the path
to an .RData-file which contains such a data set, in which case it must have the name
“Njk.i”.
It is required that this data have to be a 3-dimensional array of dimension (K+1) x (K+1) x N containing the transition counts/frequencies,
where K+1 is the number of categories k=0,...,K and N the number of objects/units/individuals.
The number of transitions (equal to time series length minus one) may be individual.
storeDirA character indicating the name of the directory (will be created if not already existing) where the results are to be stored.
priorFileIf not NULL the prior data (must have same format as mccXiPrior in LMEntryPaperData – at least the H-th entry in the list
has to be provided) or a character with the name of or the path to a file containing such data, which in this case must be named “mcc”.
The prior data contain prior information (in terms of probabilities) about transition probabilities
(possibly from another estimation procedure). For further information see Section Prior Data and mccXiPrior in LMEntryPaperData.
XThe matrix of covariates (with N rows) including the unit vector for the intercept to be included in the multinomial logit model extension.
Prior contains (see also Section Prior Data):
H An integer >= 1 indicating the number of clusters/groups.
e0 A numerical value determining the value of the prior parameter of the Dirichlet-prior for the group sizes
η_h (e0 =α_1=…=α_H, thus equal for all h).
c,cOff are necessary to calculate the prior parameter matrix for ξ (equal for all groups): diag(c) + cOff. Only used when usePriorFile=FALSE – see below.
usePriorFile If usePriorFile=TRUE, prior information for the transition probabilities as provided by priorFile are used as prior parameters for the
estimation process. In this case there are two further options depending on the value of xiPooled:
If xiPooled=TRUE, equal apriori transition probabilities are used for all groups (using
ceiling(Prior$N0*mcc[[1]]$xi)) and if xiPooled=FALSE group-specific apriori transition probabilities are used
(using ceiling(Prior$N0*mcc[[H]]$xi)).
If usePriorFile=FALSE, a priori transition probabilities are determined depending on c and cOff. In this case the
diagonal elements are set to c + cOff and the off-diagonal elements to cOff, equal for all groups.
xiPooled Only used if usePriorFile=TRUE (see above): if xiPooled=TRUE equal apriori transition probabilities are used for all groups
(using ceiling(Prior$N0*mcc[[1]]$xi)) and if xiPooled=FALSE group-specific apriori transition probabilities are used
(using
ceiling(Prior$N0*mcc[[H]]$xi)).
N0A numerical value determining a parameter for use in calculating the prior parameter matrix for ξ (see usePriorFile).
betaPrior A character. If "uninformative" (improper) prior parameters are used for the regression coefficients (i.e. betaPriorVar = ∞).
Otherwise mean and variance of the normal prior distribution for the regression coefficients have to be specified.
betaPriorMean, betaPriorVar Numerical values specifying the parameters of the normal prior distribution for the regression coefficients, only if betaPrior!="uninformative".
Initial contains:
xi.start.ind An integer taking a value out of 1, 2, 3 or 4 to determine how to define the start values for ξ:
If xi.start.ind = 1: the uniform distribution is used, meaning that all elements are equal to 1/(K+1) in all groups.
If xi.start.ind = 2: the empirical distribution/transition matrix (classical ML estimate of the transition matrix) is used (equal for all groups).
If xi.start.ind = 3: a ‘persistence’ distribution is used, meaning that the diagonal elements are equal to pers whereas all
off-diagonal elements are equal to (1-pers)/K (equal for all groups).
If xi.start.ind = 4: entry in prior file mcc[[H]]$xi is used directly for initial values.
pers Only used if xi.start.ind = 3: A numerical value (between 0 and 1) which indicates the persistence probabilities
(equal for all diagonal elements). Note, that 1/(K+1) corresponds to the uniform distribution in each row.
S.i.start A vector of length N giving an initial allocation (mandatory for mcClustExtended).
Beta.start A matrix of dimension ncol(X) x H giving start values for the regression coefficients including the zero vector in the first column
representing the baseline group.
Mcmc contains:
MAn integer indicating the overall number of iterations.
M0An integer indicating the number of the first iteration after the burn-in phase.
mOutAn integer indicating that after each mOut-th iteration a report line is written to the output window/screen.
mSaveAn integer indicating that after each mSave-th iteration an intermediate storage of the workspace is carried out.
seedAn integer indicating a random seed.
Value
A list containing (/the output file contains):
workspaceFile
A character indicating the name of and the path (based on the currend working directory) to the output file, wherein all the
results are saved.
The name of the output file starts with “MCC_” or “MCC_Logit_newAux_” respectively followed by the number of groups H, the
number of iterations M and the particular point in time when the function was called, with format: yyyymmdd_hhmmss.
E.g. MCC_H4_M10000_20110218_045254.RData or
MCC_Logit_newAux_H4_M10000_20111121_165723.RData.
Data
The argument Data.
Prior
The argument Prior.
Initial
The argument Initial.
Mcmc
The argument Mcmc.
Beta.m
A 3-dimensional array of dimension ncol(X) x H x M containing the draws for the regression coefficients
β_h in each m-th iteration step.
bk0
The prior parameters for the mean vectors of the normal (prior) distributions of the regression coefficients.
Bk0inv
The prior parameters for the inverse variance-covariance matrices of the normal (prior) distributions of the regression coefficients.
c0
A 3-dimensional array with dimension (K+1) x (K+1) x H that contains the finally used a priori parameter values
for ξ_h.
estTransProb
A 3-dimensional array with dimension (K+1) x (K+1) x H that contains the ergodic average of ξ_h for all groups
(using draws from M0 to M without thinning parameter).
fileName
A character value indicating the name of the output file (see also workspaceFile).
freq
matrix-matching (pattern recognition): a numerical vector containing the frequencies of different (!) transition matrices. (in ascending order)
indizes
matrix-matching (pattern recognition): a numerical vector containing the indices of different (!) transition matrices.
K
An integer indicating the number of categories minus one (!). See Note.
mcc
The prior data (see Section Prior Data) provided with priorFile, NULL otherwise.
N
An integer indicating N, the number of individuals/units/objects.
Njk.i
The data (see Details) provided with dataFile.
Njk.i.ind
matrix-matching (pattern recognition): the resulting Njk.i after matrix-matching.
R
matrix-matching (pattern recognition): number of different (!) transition matrices.
S.i.counts
A N x H-matrix containing the frequencies how often individual i was allocated to a certain group during the
iterations from M0+1 to codeM.
totalTime
A numeric value indicating the total time (in secs) used for the function call.
xi.hat
A matrix with dimension (K+1) x (K+1) containing the empirical transition probabilities (overall relative transition freqs).
xi.m
A 4-dimensional array of dimension M x (K+1) x (K+1) x H containing the draws for ξ_h
in each m-th iteration step.
xi.start
A matrix of dimension (K+1) x (K+1) that contains the starting values for ξ_h (only if xi.start.ind = 3).
xi.start.ind
An integer indicating the used method to calculate/determine the starting values for ξ_h.
bkN
The posterior parameters (in the last iteration step) for the mean vectors of the normal (posterior) distributions from which the regression coefficients were drawn.
BkN
The posterior parameters (in the last iteration step) for the variance-covariance matrices of the normal (posterior) distributions from which the regression coefficients were drawn.
logLike
A vector containing the values of the log-likelihood calculated in each iteration step.
logBetaPrior
A vector containing the values of the prior distribution for the regression coefficients calculated in each iteration step.
logXiPrior
A vector containing the values of the prior distribution for the transition matrices calculated in each iteration step.
logPostDens
A vector containing the values of the posterior density calculated in each iteration step.
mMax
An integer giving the position (number of iteration) of the maximum value in the posterior density logPostDens.
logClassLike
A vector containing the values of the log classification likelihood calculated in each iteration step.
entropy
A vector containing the values of the entropy calculated in each iteration step.
eta.start
Either a numeric value equal to 1/H or if xi.start.ind = 4 the corresponding data (vector) from the prior file.
estGroupSize
A numerical vector containing the ergodic average of η_h for all groups
(using draws from M0+1 to codeM without thinning parameter).
eta.m
A matrix of dimension H x M containing the draws for η_h in each m-th iteration step.
logEtaPrior
A vector containing the values of the prior distribution for the mixing proportions (group sizes) calculated in each iteration step.
Prior Data
The prior data (called mcc in the following) – to be passed via priorFile in argument-list
Data – has to be a list of lists, indexed by 1,…,H,H+1,…. Note that,
depending on parameter H (the number of groups – to be passed via H in argument-list Prior),
there have to be at least H entries (each a list). See mccXiPrior in LMEntryPaperData for
example. Within a call to dmClustering or mcClustering, at least mcc[[H]] has to
be provided as a list containing eta and xi. eta is a vector of length H containing prior
information about the relative group sizes of group h=1,…,H. xi is a 3-dimensional
array of dimension (K+1) x (K+1) x H, containing prior information in terms of
probabilities about the transition probabilities of group h=1,…,H (see examples).
Reporting Progress (Log Protocol)
The log protocol keeps record of the progress of the estimation procedure and is shown on the screen. At first the
name of the workspace file is documented. Then for each mOut-th iteration step (at least for m=1,
…, 5, 10, 20, 50, 100, 200, 500) information about the elapsed time and the expected time to the end is
reported. Finally the total time is shown.
For example:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
workspaceFile: tryN50000-sample02-01\MCC_Logit_newAux_H4_M10000_20111124_155650.RData (within current working directory!)
m = 1 ; duration of iter proc so far: 13.75 sec.
m = 2 ; duration of iter proc so far: 21.59 sec., exp time to end: 3597.97 min.
m = 3 ; duration of iter proc so far: 29.48 sec., exp time to end: 2456.18 min.
m = 4 ; duration of iter proc so far: 37.36 sec., exp time to end: 2074.93 min.
m = 5 ; duration of iter proc so far: 45.25 sec., exp time to end: 1884.66 min.
m = 10 ; duration of iter proc so far: 84.94 sec., exp time to end: 1571.55 min.
m = 20 ; duration of iter proc so far: 164.5 sec., exp time to end: 1440.24 min.
m = 50 ; duration of iter proc so far: 403.08 sec., exp time to end: 1364.3 min.
m = 100 ; duration of iter proc so far: 801.15 sec., exp time to end: 1335.38 min.
m = 200 ; duration of iter proc so far: 1530.5 sec., exp time to end: 1256.32 min.
m = 400 ; duration of iter proc so far: 3074.03 sec., exp time to end: 1232.82 min.
m = 500 ; duration of iter proc so far: 3804.67 sec., exp time to end: 1207.35 min.
m = 600 ; duration of iter proc so far: 4532.04 sec., exp time to end: 1185.47 min.
m = 800 ; duration of iter proc so far: 6075.54 sec., exp time to end: 1166.06 min.
m = 1000 ; duration of iter proc so far: 7715.48 sec., exp time to end: 1158.61 min.
...
Warning
Note that there are no such things as default values (see Section Arguments)!
Note
Note that the required data files have to be provided in the current working directory and that the results (see
Section Value) are to be saved in the directory provided by storeDir within the current working
directory. Make sure that the current working directory is set appropriately before the function is called.
Note, that in contrast to the literature (see References), the numbering (labelling) of the states of the
categorical outcome variable (time series) in this package is sometimes 0,...,K (instead of
1,...,K), however, there are K+1 categories (states)!
Author(s)
Christoph Pamminger <christoph.pamminger@gmail.com>
References
Sylvia Fruehwirth-Schnatter, Christoph Pamminger, Andrea Weber and Rudolf Winter-Ebmer, (2011),
"Labor market entry and earnings dynamics: Bayesian inference using
mixtures-of-experts Markov chain clustering".
Journal of Applied Econometrics. DOI: 10.1002/jae.1249
http://onlinelibrary.wiley.com/doi/10.1002/jae.1249/abstract
Christoph Pamminger and Sylvia Fruehwirth-Schnatter, (2010),
"Model-based Clustering of Categorical Time Series".
Bayesian Analysis, Vol. 5, No. 2, pp. 345-368. DOI: 10.1214/10-BA606
http://ba.stat.cmu.edu/journal/2010/vol05/issue02/pamminger.pdf
Sylvia Fruehwirth-Schnatter and Rudolf Fruehwirth, (2010),
"Data augmentation and MCMC for binary and multinomial logit models".
In T. Kneib and G. Tutz (eds): Statistical Modelling and Regression Structures: Festschrift in Honour of Ludwig Fahrmeir.
Physica Verlag, Heidelberg, pp. 111-132. DOI: 10.1007/978-3-7908-2413-1_7
http://www.springerlink.com/content/t4h810017645wh68/.
See also: IFAS Research Paper Series 2010-48 (http://www.jku.at/ifas/content/e108280/e108491/e108471/e109880/ifas_rp48.pdf).
See Also
dmClust, dmClustExtended, MNLAuxMix,
LMEntryPaperData, MCCExampleData,
MCCExtExampleData
Examples
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
#rm(list=ls(all=TRUE))
# ==================================================================================
if ( TRUE ) {
# ==================================================================================
# set working directory
oldDir <- getwd()
curDir <- tempdir()
setwd(curDir)
if ( !file.exists("bayesMCClust-wd") ) dir.create("bayesMCClust-wd")
setwd("bayesMCClust-wd")
myOutfilesDir <- "mcClust-Example-Outfiles"
# load data
data(MCCExampleData)
# ==================================================================================
# function call
system.time(
outList <- mcClust( # parameter lists (every four) must be complete!
Data=list(dataFile=MCCExampleData$Njk.i,
storeDir=myOutfilesDir,
priorFile= NULL),
Prior=list(H=2, # sample(2:6, 1), # 4
e0=4,
c=1,
cOff=1,
usePriorFile=FALSE,
xiPooled=FALSE,
N0=5),
Initial=list(xi.start.ind=3,
pers=0.7),
Mcmc=list(M=100,
M0=20,
mOut=5,
mSave=50,
seed=sample(1:100000, 1) # 123
)
)
)
str(outList)
#outFileName
#results <- load(outFileName)
#results
#estTransProb
allocList <- calcAllocationsMCC(outList, thin=1, maxi=50) # , plotPathsForEta=TRUE
str(allocList)
myTransProbs <- calcTransProbs(outList, estGroupSize=allocList$estGroupSize, thin=1,
printXtable=FALSE, printSd=FALSE, printTogether=TRUE )
# , plotPaths=TRUE, grLabels=paste("Group", 1:Prior$H)
str(myTransProbs)
myTransList <- plotTransProbs(outList, estTransProb=myTransProbs$estTransProb,
estGroupSize=allocList$estGroupSize, class=allocList$class, plotPooled=TRUE,
plotContTable=TRUE, printContTable=TRUE, plotContPooled=TRUE)
# , grLabels=paste("Group", 1:Prior$H)
str(myTransList)
(equiDist <- calcEquiDist(outList, thin=1, maxi=50))
#, printEquiDist=TRUE, plotEquiDist=TRUE , grLabels=paste("Group", 1:Prior$H)
myLongRunDistList <- calcLongRunDist(outList,
initialStateData=MCCExampleData$initialState,
class=allocList$class, equiDist=equiDist, maxi=50)
# , printLongRunDist=TRUE, grLabels=paste("Group", 1:Prior$H)
str(myLongRunDistList)
myTypicalMembs <- plotTypicalMembers(outList, moreTypMemb=c(10,25,40,55,70,85,100),
myObsList=MCCExampleData$obsList, classProbs=allocList$classProbs) # noTypMemb=7
str(myTypicalMembs)
plotScatter(outList, thin=1, xi11=c(1,1), xi12=c(2,2), xi21=c(2,2), xi22=c(3,3),
xi31=c(1,1), xi32=c(3,3) )
mySegPower <- calcSegmentationPower(outList, classProbs=allocList$classProbs,
class=allocList$class, printXtable=TRUE, calcSharp=TRUE, printSharpXtable=TRUE )
# , grLabels=paste("Group", 1:Prior$H)
str(mySegPower)
myEntropy <- calcEntropy(outList, classProbs=allocList$classProbs,
class=allocList$class, printXtable=TRUE )
# , grLabels=paste("Group", 1:Prior$H)
myEntropy
plotLikeliPaths(outList, from=10, by=1 )
myNumEffTables <- calcNumEff( outList, thin=1, printXi=TRUE, printE=TRUE,
printBeta=TRUE, grLabels=paste("Group", 1:outList$Prior$H) )
str(myNumEffTables)
myMSCrits <- calcMSCritMCC(workDir=myOutfilesDir, myLabel="mcClust-Example", H0=4,
whatToDoList=c("approxML", "approxMCL", "postMode") )
str(myMSCrits)
setwd(oldDir)
} # end if
# ==================================================================================
# ==================================================================================
# ==================================================================================
# ==================================================================================
if ( FALSE ) {
# ==================================================================================
rm(list=ls(all=TRUE))
# set working directory
oldDir <- getwd()
curDir <- tempdir()
setwd(curDir)
if ( !file.exists("bayesMCClust-wd") ) dir.create("bayesMCClust-wd")
setwd("bayesMCClust-wd")
myOutfilesDir <- "mcClustExtended-Example-Outfiles"
# load data
data(MCCExtExampleData)
if (!is.element("MCCExtExampleData$covariates", search())) {
attach(MCCExtExampleData$covariates)
}
# ==================================================================================
groupNr <- 2 # sample(2:6, 1) # 3
# ==================================================================================
results <- kmeans( log( MCCExtExampleData$NjkiMat + 0.5 ) , groupNr, nstart=2)
# ==================================================================================
require(nnet, quietly = TRUE)
H <- groupNr
X = cbind( intercept=1, alrateBezNew, unskilled, skilled, angStart )
N <- dim(X)[1]
mX <- data.frame( cbind(group=as.factor( results$cluster ), X[,-1],
matrix(sample(1:H,H*N,replace=TRUE),N,H)) )
colnames(mX)[6:(6+groupNr-1)] <-
c( "as.1", "as.2", "as.3", "as.4", "as.5", "as.6" )[1:groupNr]
tempMNom <- multinom(group ~ alrateBezNew+ unskilled+ skilled+ angStart,
data=as.data.frame(mX))
toStartBeta <- t(rbind(0,coef( tempMNom )))
# ==================================================================================
# function call
outList <- mcClustExtended(
Data=list(dataFile=MCCExtExampleData$Njk.i, # parameter lists must be complete!!!
storeDir=myOutfilesDir,
priorFile= NULL,
X = cbind( intercept=1, alrateBezNew, unskilled, skilled, angStart ) ),
Prior=list(H=groupNr,
c=1,
cOff=1,
usePriorFile=FALSE,
xiPooled=FALSE,
N0=5,
betaPrior = "informative", # N(0,1)
betaPriorMean = 0,
betaPriorVar = 1),
Initial=list(xi.start.ind=3,
pers=0.7,
S.i.start = results$cluster,
Beta.start = toStartBeta ),
Mcmc=list(M=100,
M0=50,
mOut=10,
mSave=50,
seed=sample(1:100000, 1) # 69814651
)
)
str(outList)
#outFileName <- outList$workspaceFile
#results <- load(outFileName)
#results
#estTransProb
allocList <- calcAllocationsMCCExt(outList, thin=1, maxi=50)
str(allocList)
myTransProbs <- calcTransProbs(outList, estGroupSize=allocList$estGroupSize, thin=1,
printXtable=FALSE, printSd=FALSE, printTogether=TRUE )
# plotPaths=TRUE, grLabels=paste("Group", 1:Prior$H)
str(myTransProbs)
myTransList <- plotTransProbs(outList, estTransProb=myTransProbs$estTransProb,
estGroupSize=allocList$estGroupSize, class=allocList$class, plotPooled=TRUE,
plotContTable=TRUE, printContTable=TRUE, plotContPooled=TRUE)
# , grLabels=paste("Group", 1:Prior$H)
str(myTransList)
(equiDist <- calcEquiDist(outList, thin=1, maxi=50))
# , printEquiDist=TRUE, plotEquiDist=TRUE, grLabels=paste("Group", 1:Prior$H)
myRegCoeffs <- calcRegCoeffs(outList, hBase=2, thin=1)
#, M0=Mcmc$M0, grLabels=paste("Group", 1:Prior$H),
# printHPD=TRUE, plotPaths=TRUE, plotACFs=TRUE
str(myRegCoeffs)
myLongRunDistList <- calcLongRunDist(outList, initialStateData=initialState,
class=allocList$class, equiDist=equiDist, maxi=50)
# , printLongRunDist=TRUE
str(myLongRunDistList)
myTypicalMembs <- plotTypicalMembers(outList, myObsList=MCCExtExampleData$obsList,
classProbs=allocList$classProbs)
# , noTypMemb=7, moreTypMemb=c(10,25,50,100,200,500,1000)
str(myTypicalMembs)
plotScatter(outList, thin=1, xi11=c(1,1), xi12=c(2,2), xi21=c(2,2), xi22=c(3,3),
xi31=c(1,1), xi32=c(3,3) )
mySegPower <- calcSegmentationPower(outList, classProbs=allocList$classProbs,
class=allocList$class, printXtable=TRUE, calcSharp=TRUE, printSharpXtable=TRUE )
# , grLabels=paste("Group", 1:Prior$H)
str(mySegPower)
myEntropy <- calcEntropy(outList, classProbs=allocList$classProbs,
class=allocList$class, printXtable=TRUE )
# , grLabels=paste("Group", 1:Prior$H)
myEntropy
plotLikeliPaths(outList, from=10, by=1 )
myNumEffTables <- calcNumEff( outList, thin=1, printXi=TRUE, printE=TRUE,
printBeta=TRUE, grLabels=paste("Group", 1:outList$Prior$H) )
str(myNumEffTables)
myMSCrits <- calcMSCritMCCExt(workDir=myOutfilesDir, NN=outList$N,
myLabel="mcClustExtended-Example", ISdraws=100, H0=3,
whatToDoList=c("approxML", "approxMCL", "postMode" ) )
str(myMSCrits)
setwd(oldDir)
# ==================================================================================
if (is.element("MCCExtExampleData$covariates", search())) {
detach(MCCExtExampleData$covariates)
}
# ==================================================================================
} # end if
# ==================================================================================
# ==================================================================================
# ==================================================================================
bayesMCClust documentation built on May 29, 2017, 3:31 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Description Usage Arguments Details Value Prior Data Reporting Progress (Log Protocol) Warning Note Author(s) References See Also Examples
Description
This function provides Markov chain clustering with or without multinomial logit model (mixtures-of-experts)
extension (see References). That is an MCMC sampler for the mixtures-of-experts extension of Markov chain
clustering. It requires four mandatory arguments: Data, Prior, Initial and Mcmc; each
representing a list of (mandatory) arguments: Data contains data information, Prior contains prior
information, Initial contains information about starting conditions (initial values) and Mcmc contains
the setup for the MCMC sampler.
Usage
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 | mcClust(
Data = list(
dataFile = stop(
"'dataFile' (=> Njk.i) must be specified: either 'filename' (path) or data"),
storeDir = "try01", priorFile = NULL),
Prior = list( H = 4, e0 = 4, c = 1, cOff = 1, usePriorFile = FALSE,
xiPooled = FALSE, N0 = 5),
Initial = list( xi.start.ind = 3, pers = 0.7, S.i.start = NULL),
Mcmc = list( M = 50, M0 = 20, mOut = 5, mSave = 10, seed = 12345))
mcClustExtended(
Data = list(
dataFile = stop(
"'dataFile' (=> Njk.i) must be specified: either 'filename' (path) or data"),
storeDir = "try01", priorFile = NULL,
X = stop("X (matrix of covariates) must be specified")),
Prior = list( H = 4, c = 1, cOff = 1, usePriorFile = FALSE,
xiPooled = FALSE, N0 = 5, betaPrior = "informative",
betaPriorMean = 0, betaPriorVar = 1),
Initial = list( xi.start.ind = 3, pers = 0.7,
S.i.start = rep(1:H, N), Beta.start = NULL),
Mcmc = list( M = 50, M0 = 20, mOut = 5, mSave = 10,
seed = 12345))
|
Arguments
Data |
a list consisting of: |
Prior |
a list consisting of: |
Initial |
a list consisting of: |
Mcmc |
a list consisting of: |
Details
Note that the values of the arguments indicated here have nothing to do with default values! For a call of these functions this lists-of-arguments structure requires a complete specification of all arguments!
The following arguments which are lists have to be completely provided (note that there are no such things as default values within lists!):
Data contains:
dataFileA 3-dim array having the transition counts/frequencies structure (like
Njk.iin the example data sets) already loaded into the current environment/workspace. Or a character with the name of or the path to an .RData-file which contains such a data set, in which case it must have the name “Njk.i”.It is required that this data have to be a 3-dimensional array of dimension (K+1) x (K+1) x N containing the transition counts/frequencies, where K+1 is the number of categories k=0,...,K and N the number of objects/units/individuals. The number of transitions (equal to time series length minus one) may be individual.
storeDirA character indicating the name of the directory (will be created if not already existing) where the results are to be stored.
priorFileIf not
NULLthe prior data (must have same format asmccXiPriorinLMEntryPaperData– at least the H-th entry in the list has to be provided) or a character with the name of or the path to a file containing such data, which in this case must be named “mcc”. The prior data contain prior information (in terms of probabilities) about transition probabilities (possibly from another estimation procedure). For further information see Section Prior Data andmccXiPriorinLMEntryPaperData.XThe matrix of covariates (with N rows) including the unit vector for the intercept to be included in the multinomial logit model extension.
Prior contains (see also Section Prior Data):
HAn integer >= 1 indicating the number of clusters/groups.
e0A numerical value determining the value of the prior parameter of the Dirichlet-prior for the group sizes η_h (
e0=α_1=…=α_H, thus equal for all h).c,cOffare necessary to calculate the prior parameter matrix for ξ (equal for all groups):
diag(c) + cOff. Only used whenusePriorFile=FALSE– see below.usePriorFileIf
usePriorFile=TRUE, prior information for the transition probabilities as provided bypriorFileare used as prior parameters for the estimation process. In this case there are two further options depending on the value ofxiPooled: IfxiPooled=TRUE, equal apriori transition probabilities are used for all groups (usingceiling(Prior$N0*mcc[[1]]$xi)) and ifxiPooled=FALSEgroup-specific apriori transition probabilities are used (usingceiling(Prior$N0*mcc[[H]]$xi)).If
usePriorFile=FALSE, a priori transition probabilities are determined depending oncandcOff. In this case the diagonal elements are set toc + cOffand the off-diagonal elements tocOff, equal for all groups.xiPooledOnly used if
usePriorFile=TRUE(see above): ifxiPooled=TRUEequal apriori transition probabilities are used for all groups (usingceiling(Prior$N0*mcc[[1]]$xi)) and ifxiPooled=FALSEgroup-specific apriori transition probabilities are used (using
ceiling(Prior$N0*mcc[[H]]$xi)).N0A numerical value determining a parameter for use in calculating the prior parameter matrix for ξ (see
usePriorFile).betaPriorA character. If "uninformative" (improper) prior parameters are used for the regression coefficients (i.e. betaPriorVar = ∞). Otherwise mean and variance of the normal prior distribution for the regression coefficients have to be specified.
betaPriorMean, betaPriorVarNumerical values specifying the parameters of the normal prior distribution for the regression coefficients, only if
betaPrior!="uninformative".
Initial contains:
xi.start.indAn integer taking a value out of 1, 2, 3 or 4 to determine how to define the start values for ξ: If
xi.start.ind = 1: the uniform distribution is used, meaning that all elements are equal to 1/(K+1) in all groups. Ifxi.start.ind = 2: the empirical distribution/transition matrix (classical ML estimate of the transition matrix) is used (equal for all groups). Ifxi.start.ind = 3: a ‘persistence’ distribution is used, meaning that the diagonal elements are equal toperswhereas all off-diagonal elements are equal to (1-pers)/K (equal for all groups). Ifxi.start.ind = 4: entry in prior filemcc[[H]]$xiis used directly for initial values.persOnly used if
xi.start.ind = 3: A numerical value (between 0 and 1) which indicates the persistence probabilities (equal for all diagonal elements). Note, that 1/(K+1) corresponds to the uniform distribution in each row.S.i.startA vector of length N giving an initial allocation (mandatory for
mcClustExtended).Beta.startA matrix of dimension
ncol(X)xHgiving start values for the regression coefficients including the zero vector in the first column representing the baseline group.
Mcmc contains:
MAn integer indicating the overall number of iterations.
M0An integer indicating the number of the first iteration after the burn-in phase.
mOutAn integer indicating that after each
mOut-th iteration a report line is written to the output window/screen.mSaveAn integer indicating that after each
mSave-th iteration an intermediate storage of the workspace is carried out.seedAn integer indicating a random seed.
Value
A list containing (/the output file contains):
workspaceFile |
A character indicating the name of and the path (based on the currend working directory) to the output file, wherein all the
results are saved.
The name of the output file starts with “ |
Data |
The argument |
Prior |
The argument |
Initial |
The argument |
Mcmc |
The argument |
Beta.m |
A 3-dimensional array of dimension ncol(X) x H x M containing the draws for the regression coefficients β_h in each m-th iteration step. |
bk0 |
The prior parameters for the mean vectors of the normal (prior) distributions of the regression coefficients. |
Bk0inv |
The prior parameters for the inverse variance-covariance matrices of the normal (prior) distributions of the regression coefficients. |
c0 |
A 3-dimensional array with dimension (K+1) x (K+1) x H that contains the finally used a priori parameter values for ξ_h. |
estTransProb |
A 3-dimensional array with dimension (K+1) x (K+1) x H that contains the ergodic average of ξ_h for all groups
(using draws from |
fileName |
A character value indicating the name of the output file (see also |
freq |
matrix-matching (pattern recognition): a numerical vector containing the frequencies of different (!) transition matrices. (in ascending order) |
indizes |
matrix-matching (pattern recognition): a numerical vector containing the indices of different (!) transition matrices. |
K |
An integer indicating the number of categories minus one (!). See Note. |
mcc |
The prior data (see Section Prior Data) provided with |
N |
An integer indicating N, the number of individuals/units/objects. |
Njk.i |
The data (see Details) provided with |
Njk.i.ind |
matrix-matching (pattern recognition): the resulting |
R |
matrix-matching (pattern recognition): number of different (!) transition matrices. |
S.i.counts |
A N x H-matrix containing the frequencies how often individual i was allocated to a certain group during the
iterations from |
totalTime |
A numeric value indicating the total time (in secs) used for the function call. |
xi.hat |
A matrix with dimension (K+1) x (K+1) containing the empirical transition probabilities (overall relative transition freqs). |
xi.m |
A 4-dimensional array of dimension M x (K+1) x (K+1) x H containing the draws for ξ_h in each m-th iteration step. |
xi.start |
A matrix of dimension (K+1) x (K+1) that contains the starting values for ξ_h (only if |
xi.start.ind |
An integer indicating the used method to calculate/determine the starting values for ξ_h. |
bkN |
The posterior parameters (in the last iteration step) for the mean vectors of the normal (posterior) distributions from which the regression coefficients were drawn. |
BkN |
The posterior parameters (in the last iteration step) for the variance-covariance matrices of the normal (posterior) distributions from which the regression coefficients were drawn. |
logLike |
A vector containing the values of the log-likelihood calculated in each iteration step. |
logBetaPrior |
A vector containing the values of the prior distribution for the regression coefficients calculated in each iteration step. |
logXiPrior |
A vector containing the values of the prior distribution for the transition matrices calculated in each iteration step. |
logPostDens |
A vector containing the values of the posterior density calculated in each iteration step. |
mMax |
An integer giving the position (number of iteration) of the maximum value in the posterior density |
logClassLike |
A vector containing the values of the log classification likelihood calculated in each iteration step. |
entropy |
A vector containing the values of the entropy calculated in each iteration step. |
eta.start |
Either a numeric value equal to 1/H or if |
estGroupSize |
A numerical vector containing the ergodic average of η_h for all groups
(using draws from |
eta.m |
A matrix of dimension H x M containing the draws for η_h in each m-th iteration step. |
logEtaPrior |
A vector containing the values of the prior distribution for the mixing proportions (group sizes) calculated in each iteration step. |
Prior Data
The prior data (called mcc in the following) – to be passed via priorFile in argument-list
Data – has to be a list of lists, indexed by 1,…,H,H+1,…. Note that,
depending on parameter H (the number of groups – to be passed via H in argument-list Prior),
there have to be at least H entries (each a list). See mccXiPrior in LMEntryPaperData for
example. Within a call to dmClustering or mcClustering, at least mcc[[H]] has to
be provided as a list containing eta and xi. eta is a vector of length H containing prior
information about the relative group sizes of group h=1,…,H. xi is a 3-dimensional
array of dimension (K+1) x (K+1) x H, containing prior information in terms of
probabilities about the transition probabilities of group h=1,…,H (see examples).
Reporting Progress (Log Protocol)
The log protocol keeps record of the progress of the estimation procedure and is shown on the screen. At first the
name of the workspace file is documented. Then for each mOut-th iteration step (at least for m=1,
…, 5, 10, 20, 50, 100, 200, 500) information about the elapsed time and the expected time to the end is
reported. Finally the total time is shown.
For example:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 | workspaceFile: tryN50000-sample02-01\MCC_Logit_newAux_H4_M10000_20111124_155650.RData (within current working directory!)
m = 1 ; duration of iter proc so far: 13.75 sec.
m = 2 ; duration of iter proc so far: 21.59 sec., exp time to end: 3597.97 min.
m = 3 ; duration of iter proc so far: 29.48 sec., exp time to end: 2456.18 min.
m = 4 ; duration of iter proc so far: 37.36 sec., exp time to end: 2074.93 min.
m = 5 ; duration of iter proc so far: 45.25 sec., exp time to end: 1884.66 min.
m = 10 ; duration of iter proc so far: 84.94 sec., exp time to end: 1571.55 min.
m = 20 ; duration of iter proc so far: 164.5 sec., exp time to end: 1440.24 min.
m = 50 ; duration of iter proc so far: 403.08 sec., exp time to end: 1364.3 min.
m = 100 ; duration of iter proc so far: 801.15 sec., exp time to end: 1335.38 min.
m = 200 ; duration of iter proc so far: 1530.5 sec., exp time to end: 1256.32 min.
m = 400 ; duration of iter proc so far: 3074.03 sec., exp time to end: 1232.82 min.
m = 500 ; duration of iter proc so far: 3804.67 sec., exp time to end: 1207.35 min.
m = 600 ; duration of iter proc so far: 4532.04 sec., exp time to end: 1185.47 min.
m = 800 ; duration of iter proc so far: 6075.54 sec., exp time to end: 1166.06 min.
m = 1000 ; duration of iter proc so far: 7715.48 sec., exp time to end: 1158.61 min.
...
|
Warning
Note that there are no such things as default values (see Section Arguments)!
Note
Note that the required data files have to be provided in the current working directory and that the results (see
Section Value) are to be saved in the directory provided by storeDir within the current working
directory. Make sure that the current working directory is set appropriately before the function is called.
Note, that in contrast to the literature (see References), the numbering (labelling) of the states of the categorical outcome variable (time series) in this package is sometimes 0,...,K (instead of 1,...,K), however, there are K+1 categories (states)!
Author(s)
Christoph Pamminger <christoph.pamminger@gmail.com>
References
Sylvia Fruehwirth-Schnatter, Christoph Pamminger, Andrea Weber and Rudolf Winter-Ebmer, (2011), "Labor market entry and earnings dynamics: Bayesian inference using mixtures-of-experts Markov chain clustering". Journal of Applied Econometrics. DOI: 10.1002/jae.1249 http://onlinelibrary.wiley.com/doi/10.1002/jae.1249/abstract
Christoph Pamminger and Sylvia Fruehwirth-Schnatter, (2010), "Model-based Clustering of Categorical Time Series". Bayesian Analysis, Vol. 5, No. 2, pp. 345-368. DOI: 10.1214/10-BA606 http://ba.stat.cmu.edu/journal/2010/vol05/issue02/pamminger.pdf
Sylvia Fruehwirth-Schnatter and Rudolf Fruehwirth, (2010), "Data augmentation and MCMC for binary and multinomial logit models". In T. Kneib and G. Tutz (eds): Statistical Modelling and Regression Structures: Festschrift in Honour of Ludwig Fahrmeir. Physica Verlag, Heidelberg, pp. 111-132. DOI: 10.1007/978-3-7908-2413-1_7 http://www.springerlink.com/content/t4h810017645wh68/. See also: IFAS Research Paper Series 2010-48 (http://www.jku.at/ifas/content/e108280/e108491/e108471/e109880/ifas_rp48.pdf).
See Also
dmClust, dmClustExtended, MNLAuxMix,
LMEntryPaperData, MCCExampleData,
MCCExtExampleData
Examples
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 | #rm(list=ls(all=TRUE))
# ==================================================================================
if ( TRUE ) {
# ==================================================================================
# set working directory
oldDir <- getwd()
curDir <- tempdir()
setwd(curDir)
if ( !file.exists("bayesMCClust-wd") ) dir.create("bayesMCClust-wd")
setwd("bayesMCClust-wd")
myOutfilesDir <- "mcClust-Example-Outfiles"
# load data
data(MCCExampleData)
# ==================================================================================
# function call
system.time(
outList <- mcClust( # parameter lists (every four) must be complete!
Data=list(dataFile=MCCExampleData$Njk.i,
storeDir=myOutfilesDir,
priorFile= NULL),
Prior=list(H=2, # sample(2:6, 1), # 4
e0=4,
c=1,
cOff=1,
usePriorFile=FALSE,
xiPooled=FALSE,
N0=5),
Initial=list(xi.start.ind=3,
pers=0.7),
Mcmc=list(M=100,
M0=20,
mOut=5,
mSave=50,
seed=sample(1:100000, 1) # 123
)
)
)
str(outList)
#outFileName
#results <- load(outFileName)
#results
#estTransProb
allocList <- calcAllocationsMCC(outList, thin=1, maxi=50) # , plotPathsForEta=TRUE
str(allocList)
myTransProbs <- calcTransProbs(outList, estGroupSize=allocList$estGroupSize, thin=1,
printXtable=FALSE, printSd=FALSE, printTogether=TRUE )
# , plotPaths=TRUE, grLabels=paste("Group", 1:Prior$H)
str(myTransProbs)
myTransList <- plotTransProbs(outList, estTransProb=myTransProbs$estTransProb,
estGroupSize=allocList$estGroupSize, class=allocList$class, plotPooled=TRUE,
plotContTable=TRUE, printContTable=TRUE, plotContPooled=TRUE)
# , grLabels=paste("Group", 1:Prior$H)
str(myTransList)
(equiDist <- calcEquiDist(outList, thin=1, maxi=50))
#, printEquiDist=TRUE, plotEquiDist=TRUE , grLabels=paste("Group", 1:Prior$H)
myLongRunDistList <- calcLongRunDist(outList,
initialStateData=MCCExampleData$initialState,
class=allocList$class, equiDist=equiDist, maxi=50)
# , printLongRunDist=TRUE, grLabels=paste("Group", 1:Prior$H)
str(myLongRunDistList)
myTypicalMembs <- plotTypicalMembers(outList, moreTypMemb=c(10,25,40,55,70,85,100),
myObsList=MCCExampleData$obsList, classProbs=allocList$classProbs) # noTypMemb=7
str(myTypicalMembs)
plotScatter(outList, thin=1, xi11=c(1,1), xi12=c(2,2), xi21=c(2,2), xi22=c(3,3),
xi31=c(1,1), xi32=c(3,3) )
mySegPower <- calcSegmentationPower(outList, classProbs=allocList$classProbs,
class=allocList$class, printXtable=TRUE, calcSharp=TRUE, printSharpXtable=TRUE )
# , grLabels=paste("Group", 1:Prior$H)
str(mySegPower)
myEntropy <- calcEntropy(outList, classProbs=allocList$classProbs,
class=allocList$class, printXtable=TRUE )
# , grLabels=paste("Group", 1:Prior$H)
myEntropy
plotLikeliPaths(outList, from=10, by=1 )
myNumEffTables <- calcNumEff( outList, thin=1, printXi=TRUE, printE=TRUE,
printBeta=TRUE, grLabels=paste("Group", 1:outList$Prior$H) )
str(myNumEffTables)
myMSCrits <- calcMSCritMCC(workDir=myOutfilesDir, myLabel="mcClust-Example", H0=4,
whatToDoList=c("approxML", "approxMCL", "postMode") )
str(myMSCrits)
setwd(oldDir)
} # end if
# ==================================================================================
# ==================================================================================
# ==================================================================================
# ==================================================================================
if ( FALSE ) {
# ==================================================================================
rm(list=ls(all=TRUE))
# set working directory
oldDir <- getwd()
curDir <- tempdir()
setwd(curDir)
if ( !file.exists("bayesMCClust-wd") ) dir.create("bayesMCClust-wd")
setwd("bayesMCClust-wd")
myOutfilesDir <- "mcClustExtended-Example-Outfiles"
# load data
data(MCCExtExampleData)
if (!is.element("MCCExtExampleData$covariates", search())) {
attach(MCCExtExampleData$covariates)
}
# ==================================================================================
groupNr <- 2 # sample(2:6, 1) # 3
# ==================================================================================
results <- kmeans( log( MCCExtExampleData$NjkiMat + 0.5 ) , groupNr, nstart=2)
# ==================================================================================
require(nnet, quietly = TRUE)
H <- groupNr
X = cbind( intercept=1, alrateBezNew, unskilled, skilled, angStart )
N <- dim(X)[1]
mX <- data.frame( cbind(group=as.factor( results$cluster ), X[,-1],
matrix(sample(1:H,H*N,replace=TRUE),N,H)) )
colnames(mX)[6:(6+groupNr-1)] <-
c( "as.1", "as.2", "as.3", "as.4", "as.5", "as.6" )[1:groupNr]
tempMNom <- multinom(group ~ alrateBezNew+ unskilled+ skilled+ angStart,
data=as.data.frame(mX))
toStartBeta <- t(rbind(0,coef( tempMNom )))
# ==================================================================================
# function call
outList <- mcClustExtended(
Data=list(dataFile=MCCExtExampleData$Njk.i, # parameter lists must be complete!!!
storeDir=myOutfilesDir,
priorFile= NULL,
X = cbind( intercept=1, alrateBezNew, unskilled, skilled, angStart ) ),
Prior=list(H=groupNr,
c=1,
cOff=1,
usePriorFile=FALSE,
xiPooled=FALSE,
N0=5,
betaPrior = "informative", # N(0,1)
betaPriorMean = 0,
betaPriorVar = 1),
Initial=list(xi.start.ind=3,
pers=0.7,
S.i.start = results$cluster,
Beta.start = toStartBeta ),
Mcmc=list(M=100,
M0=50,
mOut=10,
mSave=50,
seed=sample(1:100000, 1) # 69814651
)
)
str(outList)
#outFileName <- outList$workspaceFile
#results <- load(outFileName)
#results
#estTransProb
allocList <- calcAllocationsMCCExt(outList, thin=1, maxi=50)
str(allocList)
myTransProbs <- calcTransProbs(outList, estGroupSize=allocList$estGroupSize, thin=1,
printXtable=FALSE, printSd=FALSE, printTogether=TRUE )
# plotPaths=TRUE, grLabels=paste("Group", 1:Prior$H)
str(myTransProbs)
myTransList <- plotTransProbs(outList, estTransProb=myTransProbs$estTransProb,
estGroupSize=allocList$estGroupSize, class=allocList$class, plotPooled=TRUE,
plotContTable=TRUE, printContTable=TRUE, plotContPooled=TRUE)
# , grLabels=paste("Group", 1:Prior$H)
str(myTransList)
(equiDist <- calcEquiDist(outList, thin=1, maxi=50))
# , printEquiDist=TRUE, plotEquiDist=TRUE, grLabels=paste("Group", 1:Prior$H)
myRegCoeffs <- calcRegCoeffs(outList, hBase=2, thin=1)
#, M0=Mcmc$M0, grLabels=paste("Group", 1:Prior$H),
# printHPD=TRUE, plotPaths=TRUE, plotACFs=TRUE
str(myRegCoeffs)
myLongRunDistList <- calcLongRunDist(outList, initialStateData=initialState,
class=allocList$class, equiDist=equiDist, maxi=50)
# , printLongRunDist=TRUE
str(myLongRunDistList)
myTypicalMembs <- plotTypicalMembers(outList, myObsList=MCCExtExampleData$obsList,
classProbs=allocList$classProbs)
# , noTypMemb=7, moreTypMemb=c(10,25,50,100,200,500,1000)
str(myTypicalMembs)
plotScatter(outList, thin=1, xi11=c(1,1), xi12=c(2,2), xi21=c(2,2), xi22=c(3,3),
xi31=c(1,1), xi32=c(3,3) )
mySegPower <- calcSegmentationPower(outList, classProbs=allocList$classProbs,
class=allocList$class, printXtable=TRUE, calcSharp=TRUE, printSharpXtable=TRUE )
# , grLabels=paste("Group", 1:Prior$H)
str(mySegPower)
myEntropy <- calcEntropy(outList, classProbs=allocList$classProbs,
class=allocList$class, printXtable=TRUE )
# , grLabels=paste("Group", 1:Prior$H)
myEntropy
plotLikeliPaths(outList, from=10, by=1 )
myNumEffTables <- calcNumEff( outList, thin=1, printXi=TRUE, printE=TRUE,
printBeta=TRUE, grLabels=paste("Group", 1:outList$Prior$H) )
str(myNumEffTables)
myMSCrits <- calcMSCritMCCExt(workDir=myOutfilesDir, NN=outList$N,
myLabel="mcClustExtended-Example", ISdraws=100, H0=3,
whatToDoList=c("approxML", "approxMCL", "postMode" ) )
str(myMSCrits)
setwd(oldDir)
# ==================================================================================
if (is.element("MCCExtExampleData$covariates", search())) {
detach(MCCExtExampleData$covariates)
}
# ==================================================================================
} # end if
# ==================================================================================
# ==================================================================================
# ==================================================================================
|
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.