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
|
#
# Copyright 2007-2020 by the individuals mentioned in the source code history
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#------------------------------------------------------------------------------
# Author: Michael D. Hunter
# Date: 2015-03-11
# Filename: MxMI.R
# Purpose: Write a function for model modification
#------------------------------------------------------------------------------
# TODO
# Get the submodel and multigroup model MI working
# Write automatic model search algorithm: 1. forward entry (start with minimal base model and add free parameters until no longer merited)
mxMI <- function(model, matrices=NA, full=TRUE){
warnModelCreatedByOldVersion(model)
if(single.na(matrices)){
matrices <- names(model$matrices) #names of them rather
if (is(model$expectation, "MxExpectationRAM")) {
matrices <- setdiff(matrices, model$expectation$F)
}
}
if(imxHasWLS(model)){stop("modification indices not implemented for WLS fitfunction")}
param <- omxGetParameters(model)
param.names <- names(param)
gmodel <- omxSetParameters(model, free=FALSE, labels=param.names)
#same as original model but with all parameters fixed
mi.r <- NULL
mi.f <- NULL
a.names <- NULL
new.models <- list()
for(amat in matrices){
matObj <- model[[amat]]
freemat <- matObj$free
sym.sel <- upper.tri(freemat, diag=TRUE)
notSymDiag <- !(is(gmodel[[amat]])[1] %in% c("DiagMatrix", "SymmMatrix"))
for(i in 1:length(freemat)){
# only walk the lower triangle of Diag and Symm matrices.
if(freemat[i]==FALSE && ( notSymDiag || sym.sel[i]==TRUE )){
tmpLab <- gmodel[[amat]]$labels[i]
plusOneParamModel <- model
if(length(tmpLab) > 0 && !is.na(tmpLab)){
gmodel <- omxSetParameters(gmodel, labels=tmpLab, free=TRUE)
plusOneParamModel <- omxSetParameters(plusOneParamModel, labels=tmpLab, free=TRUE)
} else{
#free single parameter of model that was fixed in original model
gmodel[[amat]]$free[i] <- TRUE
# create a new model with all the free params of the orig
# PLUS the new one under consideration
plusOneParamModel[[amat]]$free[i] <- TRUE
}
# specific adjustments for zero matrices
if(is(gmodel[[amat]])[1] %in% c("ZeroMatrix")){
cop <- gmodel[[amat]]
newSingleParamMat <- mxMatrix("Full", nrow=nrow(cop),
ncol=ncol(cop), values=cop$values, free=cop$free,
labels=cop$labels, name=cop$name, lbound=cop$lbound,
ubound=cop$ubound, dimnames=dimnames(cop))
bop <- plusOneParamModel[[amat]]
newPlusOneParamMat <- mxMatrix("Full", nrow=nrow(bop),
ncol=ncol(bop), values=bop$values, free=bop$free,
labels=bop$labels, name=bop$name, lbound=bop$lbound,
ubound=bop$ubound, dimnames=dimnames(bop))
# specific adjustments for symmetric and diagonal matrices
} else if(is(gmodel[[amat]])[1] %in% c("DiagMatrix", "SymmMatrix")){
cop <- gmodel[[amat]]
newSingleParamMat <- mxMatrix("Symm", nrow=nrow(cop),
ncol=ncol(cop), values=cop$values,
free=(cop$free | t(cop$free)), labels=cop$labels,
name=cop$name, lbound=cop$lbound, ubound=cop$ubound,
dimnames=dimnames(cop))
bop <- plusOneParamModel[[amat]]
newPlusOneParamMat <- mxMatrix("Symm", nrow=nrow(bop),
ncol=ncol(bop), values=bop$values,
free=(bop$free | t(bop$free)), labels=bop$labels,
name=bop$name, lbound=bop$lbound, ubound=bop$ubound,
dimnames=dimnames(bop))
# no adjustments. just fill in
} else {
newSingleParamMat <- gmodel[[amat]]
newPlusOneParamMat <- plusOneParamModel[[amat]]
}
gmodel[[amat]] <- newSingleParamMat
plusOneParamModel[[amat]] <- newPlusOneParamMat
# The custom compute plan. Only do derivatives
custom.compute <- mxComputeSequence(list(mxComputeNumericDeriv(checkGradient=FALSE),
mxComputeReportDeriv()))
# Create and run the single-parameter model for the LISREL-type/partial/[lower bound] MI
gmodel <- mxModel(gmodel, custom.compute)
grun <- try(mxRun(gmodel, silent = FALSE, suppressWarnings = FALSE, unsafe=TRUE)) #suppress War
nings =TRUE
if (is(grun, "try-error")) {
# oops, not happy about that parameter (e.g., diag of RAM's A)
gmodel <- omxSetParameters(gmodel, labels=names(omxGetParameters(gmodel)), free=FALSE)
next
}
# restricted MI
grad <- grun$output$gradient #get gradient
hess <- grun$output$hessian #get Hessian
modind <- 0.5*grad^2/hess #use 0.5*g^2/k
if(full==TRUE){
custom.compute.smart <- mxComputeSequence(list(
mxComputeNumericDeriv(knownHessian=model$output$hessian, checkGradient=FALSE),
mxComputeReportDeriv()))
# Create and run the all-plus-one-parameter model for the Mplus-type/full MI
plusOneParamRun <- mxRun(mxModel(plusOneParamModel, custom.compute.smart), silent = FALSE, suppressWarnings = FALSE, unsafe=TRUE)
# full MI
grad.full <- plusOneParamRun$output$gradient
grad.full[is.na(grad.full)] <- 0
hess.full <- plusOneParamRun$output$hessian
modind.full <- 0.5*t(matrix(grad.full)) %*% solve(hess.full) %*% matrix(grad.full)
} else {
modind.full <- NULL
}
n.names <- names(omxGetParameters(grun))
if(length(modind) > 0){
a.names <- c(a.names, n.names)
mi.r <- c(mi.r, modind)
mi.f <- c(mi.f, modind.full)
new.models <- c(new.models, plusOneParamModel)
}
gmodel <- omxSetParameters(gmodel, labels=names(omxGetParameters(gmodel)), free=FALSE)
}
}
names(mi.r) <- a.names
if(full==TRUE) {names(mi.f) <- a.names}
names(new.models) <- a.names
}
# not yet tested
if(length(model$submodels) > 0){
for(asubmodel in names(model$submodels)){
ret <- c(ret, mxMI(asubmodel)) #probably won't work.
}
}
return(list(MI=mi.r, MI.Full=mi.f, plusOneParamModels=new.models))
}
|
