-
Notifications
You must be signed in to change notification settings - Fork 3
/
ChangeLog
373 lines (289 loc) · 15.2 KB
/
ChangeLog
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
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
CHANGES IN VERSION 4.0.0
- Added fast Cholesky updates for information criteria
- Added fast Cholesky updates for modelSelection under pMOM, normalid,
icarplus and most other priors (not peMOM, piMOM though)
- Added class lmObject to organize linear / GLM objects
- Modified all regression marginal likelihood routines to operate with
Armadillo sparse matrices
- Updated modelSelectionEnum, modelSelectionGibbs, greedyVarSel to work with
models stored as Armadillo sparse matrices
CHANGES IN VERSION 3.5.5
- Added fast Cholesky updates for GGM
- Added almost parallel sampler to modelSelectionGGM
- Fixed bug in localnulltest occuring for multivariate coordinates
- Fixed memory leak in set_marginalPars
- Added biocViews: to DESCRIPTION for automatic installation of bioconductor
packages
- Fixed memory leak in GGM birth/death proposal
- Removed many importFrom instances from NAMESPACE, replace by :: in code
- Fixed bug in bf.mom for base t density
CHANGES IN VERSION 3.5.4
- Changed all .Call interfaces to Rcpp exports, to avoid warnings when turning
on -flto compiler flag
CHANGES IN VERSION 3.5.3
- Fixed memory allocation error in modselIntegralsGGM
CHANGES IN VERSION 3.5.0
- Added modelSelectionGGM
- Added prettier iteration progress
- Disable sgd initialization in modelSelection, due to discontinuation of
package sgd
CHANGES IN VERSION 3.4.3
- Fixed bug in coef method for msfit objects that affected GLMs (continuous
outcomes were not affected)
CHANGES IN VERSION 3.4.2
- Fixed bug in Laplace approximation to AFT marginal likelihood under pMOM prior
CHANGES IN VERSION 3.4.1
- Fixed minor bug in localnulltest_fda
CHANGES IN VERSION 3.4.0
- Added localnulltest
- Added approximate posterior sampling method in LM when prior not
specifically implemented
- Added check in modelSelection to detect NAs and infinite values in y or x
- Added localnulltest and localtest class
- Fixed bug in predict.msfit that could occur when colnames(object$xstd) was
null
CHANGES IN VERSION 3.3.1
- Added info on varnames and top model to the output of bestAIC, bestBIC, bestEBIC, bestIC
- Added methods for icfit objects: coef, confint, show, summary
CHANGES IN VERSION 3.3.0
- Fixed a bug in the calculation of BIC (it was off by log(n), and failed for
the null model with no covariates)
- Added bestAIC, bestBIC, bestEBIC, bestIC functions
CHANGES IN VERSION 3.2.0
- Adjusted cil so it has the same arguments as modelSelection. Added arguments priorCoef, center, scale, includevars and verbose. Removed arguments beta.prior and tau. Set the defaults of center and scale as in modelSelection (TRUE for all non-factor variables)
- Added GLM functionality to cil via arguments family, familyD which can now also be 'binomial' and 'poisson'
- Added plotprior method to cilfit
- cil internal parameter rho.min now set to 1/p rather than 1/p^2
- cil now returns BMA inference for all parameters in object$beta
- cil now returns BMA estimates, intervals and margpp in a single element object$cil.teff (before they were stored separately)
- cil now extracts BMA inference using coef function from msfit objects and allows for any priorCoef allowed by modelSelection
- Fixed incorrect BIC formula in CIL's lasso.bic for logistic regression
- cil now allows data frames as inputs
- Fixed bug in modelSelection where margpp==0 for variables with includevars==TRUE
CHANGES IN VERSION 3.1.4
- Removed pmomLM, pmomPM, emomLM, emomPM, zellnerLM, zellnerPM, pplProbit
CHANGES IN VERSION 3.1.3
- Fixed NOTE from comparing class() to string in cil.R
CHANGES IN version 3.1.2
- Added cil
- Fixed gcc warnings for comparison of pointer addition to NULL
CHANGES IN VERSION 3.1.1
- Added bicprior to linear models
CHANGES IN VERSION 3.1.0
- Fixed modelSelection to work with GLMs with bicprior() and group constraints
- Added initpar option to modelSelection with 'MLE', 'MLE-aisgd', 'L1',
'L2-aisgd' to improve ALA accuracy and increase speed of Laplace approx
CHANGES IN VERSION 3.0.7
- Fixed bug in predict.msfit where covariates in newdata were not scaled appropriately
CHANGES IN VERSION 3.0.6
- Added posterior sampling under Normal approximation for GLMs, in the
presence of groups
CHANGES IN VERSION 3.0.5
- Added getBIC and getEBIC methods
CHANGES IN VERSION 3.0.4
- Added bicprior() to modelSelection
- Fixed memory error in pmomLM Rd examples file
CHANGES IN VERSION 3.0.3
- Added step size decrease in Newton-Raphson optimization (modselfunction.cpp)
- Increased default max step size decreases for GLMs from 5 to 10
- Added optim_maxit option to modelSelection and nlpMarginal
CHANGES IN VERSION 3.0.2
- Fixed numerical overflow issue in postProb
CHANGES IN VERSION 3.0.1
- Added logistic and Poisson regression to modelSelection
- Added ALA for GLMs
CHANGES IN VERSION 2.3.2
- Added method='ALA' to modelSelection for AFT survival models
- Implemented asymetric modelbinomprior
- Fix calibration of groupmomprior
- Add taustd prior to set dispersion in terms of unit information prior
CHANGES IN VERSION 2.3.1
- Added option to specify multiple nknots in modelSelection
- Implement groupzellner prior combinations with MOM, Zellner and normalid
priors
- Implement groupmom prior with MOM
- Fix some inconsistencies between tau and taugroup usage
CHANGES IN VERSION 2.3.0
- Fixed error in coefByModel (typo in calling postProb)
- Fixed unnecessary warnings returned by hastPostSamples
- Fixed error in modelSelection Gibbs when using includevars
CHANGES IN VERSION 2.2.9
- Fixed NOTE due to specifying non-canonical URL for mombf package
CHANGES IN VERSION 2.2.8
- Fixed error in Windows checks related to a test in test-modelSelection-Gibbs.R
CHANGES IN VERSION 2.2.7
- Fixed error caused by class(smoothterms) returning multiple values
CHANGES IN VERSION 2.2.6
- Fixed error caused by class(y) where y is "matrix" returning multiple values
- Added coefByModel, and hidden function modelSelectionGLM with preliminary GLM support
- Added nomalidprior
- Added support for groups and constraints to nlpMarginal
CHANGES IN VERSION 2.2.5
- Fixed bug that prevented modelSelection to run for survival outcomes where no formula is specified
CHANGES IN VERSION 2.2.4
- Fixed bug in C++ code for checkConstraints
- Implemented Newton-Raphson optim in modselFunction.cpp
- Implemented MOM orthogonal approx in SurvMarg
- greedyVarSel now does unconstrained search, the result is subsequently
modified to satisfy constraints
- Improved modelSelection parameter initialization via Taylor expansion at 0
CHANGES IN VERSION 2.2.3
- Removed cleanup script that wrongly deleted the Makevars file
CHANGES IN VERSION 2.2.2
- Improved speed of modelSelection for AFT models (improved Mill's ratio
evaluation and avoided unnecessary calls to pnorm)
CHANGES IN VERSION 2.2.1
- Fixed bug in coef and predict methods for msfit objects
- Fixed compiler warnings in Windows
CHANGES IN VERSION 2.2.0
- Fixed bug in modelSelection when computing Binomial model prior probabilities
- Added Normal Accelerated Failure Time models
- Added coef method to extract BMA estimates and intervals from msfit objects
- Added predict method to extract BMA predictions and intervals from msfit objects
- Added additive models to modelSelection
- Added support for formulas to modelSelection
- Added RcppArmadillo
- Added option to group variables and impose hierarchical restrictions to
modelSelection
- Added rnlp method that only requires msfit
- msfit objects now store priorCoef, priorDelta, priorVar and standardized versions of (y,x)
- Added baseDensity='laplace' to dmom
- Added dmomigmarg, pmomigmarg
CHANGES IN VERSION 2.1.2
- Added complexity prior to modelSelection and bms_ortho
- Added option includeModels to postModeOrtho.
- Fixed bug in postModeOrtho under Zellner's prior that caused it to return wrong posterior model probabilities
- rnlp now accepts y formatted as a 1-column matrix
- pMOM integrated likelihood now defaults to orthogonal approximation, under method=='auto'
CHANGES IN VERSION 2.1.1
- Fixed R_registerRoutines and R_useDynamicSymbols NOTE when running R CMD check
CHANGES IN VERSION 2.1.0
- Added Normal - Inverse Wishart conjugate analysis
- Added Normal mixtures
- Removed dependency on package actuar
CHANGES IN VERSION 1.9.6
- Removed register statements causing a warning in new C++ compiler
CHANGES IN VERSION 1.9.5
- Added option includevars to modelSelection
- Added eMOM and iMOM priors for two-piece Normal and two-piece Laplace errors
- Fixed a bug in integrated likelihood Laplace approximation for the MOM prior and Normal errors
- Fixed valgrind memory warnings
CHANGES IN VERSION 1.8.3
- Fixed spectral clustering on cor(x)^2 instead of cor(x)
CHANGES IN VERSION 1.8.2
- Added blocksearch and spectral clustering
CHANGES IN VERSION 1.8.1
- Fixed use of infinity that was creating trouble in some C compilers
CHANGES IN VERSION 1.8.0
- Major vignette update
- Added model enumeration option to modelSelection
- Added integrated likelihood for product eMOM prior in linear models
- Added general nlpMarginal function
- Added linear regression with two-piece normal, laplace and two-piece laplace residuals
- Fixed numerical stability issues in orthogonal and block-diagonal regression
CHANGES IN VERSION 1.7.1
- Fixed issue in block-diagonal variable selection that caused it to skip some model sizes
CHANGES IN VERSION 1.7.0
- Added block-orthogonal and block-diagonal variable selection
CHANGES IN VERSION 1.6.1
- Fixed maximum number of variables <=n in pmomLM, pmomLM (this also solves a numerical overflow issue when p>n)
CHANGES IN VERSION 1.6.0
- Added pimomCoxMarginalR
- Added method=='plugin' to modelSelection, pimomMarginalK, pimomMarginalU
CHANGES IN VERSION 1.5.9
- Fixed warning in logPL about not found mclapply
- Fixed R functions computing pMOM, piMOM, and peMOM marginal likelihood when input model has no covariates
CHANGES IN VERSION 1.5.8
- Fixed bug in emomLM when sampling residual variance under model with no covariates
- Added rnlp method for objects of type 'Surv'
- Added rnlp method for general d(theta) * N(theta;m,V)
- Added pmomCoxMarginalR
- Added function eprod
CHANGES IN VERSION 1.5.7
- Fixed memory access error in pmomLM
CHANGES IN VERSION 1.5.6
- Fixed apparently harmless invalid memory read at modselIntegrals.cpp
CHANGES IN VERSION 1.5.5
- Implemented faster Laplace approx for piMOM marginal likelihood (now uses gradient algorithm to find posterior mode)
- Implemented p>n case in modelSelection
- Changed default to method=='auto' in modelSelection
- Changed default priorCoef, priorModel, priorVar in modelSelection to recommendations in Johnson & Rossell (2010)
- Fixed bug in modelSelection when method=='auto' and a prior different from MOM was used.
CHANGES IN VERSION 1.5.4
- Fixed R CMD CHECK warnings
CHANGES IN VERSION 1.5.3
- Fixed compiler warnings
CHANGES IN VERSION 1.5.0
- Added function rnlp to sample from MOM, eMOM and iMOM posteriors under Normal linear models
- Added class msfit and several basic methods
- Added momprior, imomprior, emomprior, modelunifprior, modelbinomprior, modelbbprior, igprior functions to facilitate creating objects of class msPriorSpec
- Prevented numerical overflow in computing iMOM integrated likelihood with method=='Hybrid' and method=='MC'. Improves pimomMarginalU and modelSelection for both these methods.
CHANGES IN VERSION 1.4.1
- Fixed bug in modelSelection that caused to report the wrong model (only) at the first Gibbs iteration
- Increased precision of Laplace approximation to compute MOM and iMOM integrated likelihood
CHANGES IN VERSION 1.4.0
- Fixed C++ bug in Solaris caused by overloading of 'log' and 'sqrt' functions
CHANGES IN VERSION 1.3.9
- Fixed C++ bug in Solaris caused by overloading of 'pow' function
- Limited number of saved models by modelSelection Gibbs scheme when p>16, to avoid running out of memory
CHANGES IN VERSION 1.3.8
- modelSelection Gibbs scheme now uses saved log(integrated likelihood) + log(prior) for previously considered models. Observed a 5-10 fold increase in speed.
- modelSelection with Beta-Binomial prior now marginalizes over Binomial success prob, which greatly improves mixing of the Gibbs sampler.
CHANGES IN VERSION 1.3.7
- Fixed warning due to visibility of mclapply
CHANGES IN VERSION 1.3.6
- Added truncated multivariate normal random number generation routines
CHANGES IN VERSION 1.3.5
- Increased precision in pimom marginal Laplace approximation, which could cause crashing due to non-positive definitiness
CHANGES IN VERSION 1.3.4
- Fixed code chunks in vignette to enhance their visibility
CHANGES IN VERSION 1.3.2
- Added check to T dmom to ensure that prior is proper
- Added greedy initialization algorithm for probit models
- Added multivariate version of heavy-tailed pMOM to dmom
- Modified heavy-tailed eMOM implemented in dmom so that it is consistent with heavy-tailed pMOM. Now both are defined via hyper-priors on tau.
- Added pplPM function
- Added emomLM, emomPM function
CHANGES IN VERSION 1.3.1
- modelSelection now returns the estimated regression coefficients for the most probable model
CHANGES IN VERSION 1.3.0
- Added pmomPM function
- Added pmomLM function
CHANGES IN VERSION 1.2.1
- Added dmom prior for unspecified tau. Added parameters a.tau and b.tau to dmom.
CHANGES IN VERSION 1.2
- Added demom and pemom functions.
CHANGES IN VERSION 1.1.3
- Improved vignette model selection section
- Implemented new Laplace approximation for product iMOM marginals which is much more efficient computationally than the previous one
CHANGES IN VERSION 1.1.2
- Removed makefile from inst/doc
- Added hierarchical Beta-Binomial prior for model space indicator in modelSelection
CHANGES IN VERSION 1.1.1
- Fixed some C warnings due to uninitialized variables
CHANGES IN VERSION 1.1.0
- Product MOM and iMOM priors now implemented
- Linear model variable selection for MOM and iMOM priors implemented
CHANGES IN VERSION 1.0.5
- Implemented product iMOM prior density in dpimom. Renamed quadratic iMOM density from dimom to dqimom. Added wrapper dimom to access either dpimom or dqimom.
CHANGES IN VERSION 1.0.4
- Implemented tMom Bayes factors as an option to mombf, mode2g and g2mode.
- Implemented tMom prior density as an option to dmom
CHANGES IN VERSION 1.0.3
- A couple minor changes in documentation files .Rd to avoid warnings
CHANGES IN VERSION 1.0.2
- Fixed bug in the multivariate calculation of MOM Bayes Factors (affects mombf, momknown)
CHANGES IN VERSION 1.0.1
- Fixed bug in the multivariate calculation of dmom
CHANGES IN VERSION 1.0.0
- Added option 'method' to functions 'imombf', 'momknown' and 'imomunknown' to allow users to choose integration method. Default integration method changed from Monte Carlo to integration based on routine 'integrate' from package 'stats'.
- Functions 'imombf', 'imomknown' and 'imomunknown' now accept parameter 'g' to be a vector instead of a single value.
- Added pmom and pimom functions
- Removed g2mode.univ and mode2g.univ functions. The same results can now be obtained via g2mode and mode2g.
- Corrected error in dimom evaluation
- Corrected error in zbfknown evaluation
- Changed order of dmom and dimom arguments: theta0 is not 1st argument anymore
- Added contents to vignette illustrating the new functions, and changed the name of the vignette from "mombf_manual" to "mombf"
CHANGES IN VERSION 0.0.1
- Changed word "library" in Vignette title to "package"
- Indicated GPL(>=2) in the DESCRIPTION file