LTPDA Toolbox™

Method matrix/modelSelect

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
    modelselect.m - method to compute the Bayes Factor using
                    RJMCMC, LF, LM, SBIC methods
 
             call - Bxy = modelselect(out,pl)
 
           inputs - out: matrix objects with measured outputs
                    pl:  parameter list
 
          outputs - Bxy:
                    -RJMCMC:
                      an array of AOs containing the evolution
                      of the Bayes factors. (comparing each model
                      with each other)
 
                    -LM, LF, SBIC:
                      An ao containing the Bayes Factor for the
                      comparison of 2 models
 
 Parameters Description
 
    N. Karnesis 27/09/2011
 
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Method Details
Access	public
Defining Class	matrix
Sealed	0
Static	0

Parameter Description

Sets for this method …
Default

Default
no description
Key	Default Value	Options	Description
modelSelect
METHOD	'RJMCMC'	'RJMCMC' 'LF' 'LM' 'SBIC'	Method to use to compare two (or more) models. The choises available are; RJMCMC, Laplace-Metropolis (LM), Laplace-Fisher (LF) and Schwarz-Bayes Information Criterion (SBIC). Default is RJMCMC.Each method needs some compulsory fields to be filled in the plist.
NEFF	[]	none	Number of effective samples to be used for the estimation of the SBIC. If left empty, then Neff = number of samples of the data.
THETAMAP	[]	none	Cell array containing the parameter vectors that maximize the likelihoods for two models. Used for LF, LM, SBIC methods.
SAMPLE	1	1 0	Used in the case of 'LM' method to extract covariance matrices. If true, MCMC parameter estimationis applied for both models. If false, the covariance matrices and parameter values in the 'cov' and 'thetaMAP' fields are used to calculate the evidence.
LOG PARAMETERS	''	none	Select the parameters to be treated in log scale.
mve
PCA	0	0 1	Set to true to perform Principal Component Analysis.
M	100	none	Number 'm' of random sub-samples to be drawn from the data. If set to zero, the method will attempt to proceed taking into acount all possible sub-samples. ATTENTION: If the data-set is too large, this computation is practically unfeasible!
DISCARD	0	none	Discard the first number of samples.
rjsample
INNAMES	''	none	Input names. Used for ssm models
OUTNAMES	''	none	Output names. Used for ssm models
MODELS	''	none	A cell array input of models.
FITPARAMS	[]	none	A cell array of evaluated parameters for each model.
INPUT	''	none	A matrix array of input signals.
N	1000	none	number of samples of the chain.
COV	0.0001	none	Cell array containing the covariances of the gaussian jumping distribution for each model.
NOISE	''	none	A matrix array of noise spectrum (PSD) used to compute the likelihood.
MODELFREQDEPENDENT	1	1 0	Set to true to use frequency dependent s models, set to false when using constant models
SEARCH	1	1 0	Set to true to use bigger jumps in parameter space during annealing and cool down.
FREQUENCIES	[]	none	Range of frequencies where the analysis is performed. If an array, only first and last are used
FSOUT	[]	none	Desired sampling frequency to resample the input time series
HEAT	1	none	The heat index flattening likelihood surface during annealing.
TC	''	none	An array of two values setting the initial and final value for the cooling down.
X0	[]	none	The proposed initial values (cell array again).
RANGE	[]	none	Ranges
BURNIN	1	none	If method is RJMCMC, choose number of samples to be discarded to compute B.
JUMPS	[]	none	An array of four numbers setting the rescaling of the covariance matrix during the search phase.The first value is the one applied by default, the following thhree apply just when the chain sample ismod(10), mod(25) and mod(100) respectively.
PLOT BF	[]	none	Case: RJMCMC: If set equal to true, the evolution of the Bayes factor and the Loglikelihoodsare plotted every 500 steps. Case: LM: vector for plotting the chains during the mcmc runs.
PLOT PARAMETERS	[]	none	A cell array that includes the parameters names desired to create the trace plots.
DEBUG	0	0 1	Set to true to get debug information of the MCMC process.
OUTMODEL	''	none	Output model. Still under test
SCALE MATRIX	''	none	A matrix array of noise spectrum (PSD) used to compute the likelihood. It is possible to input just a scale matrix, containing the desirable weights.
LOG-LIKELIHOOD	''	none	The log-likelihood to sample with the MH algorithm. If left empty, then the standard Gaussian approximation will be employed.
FREQUENCIES VECTOR	200	none	A vector of frequencies. Used for the update of the Fisher Matrix during the MH sampling.
INMODEL	''	none	Input model. Still under test
FPRINT	100	none	Print progress on screen every specified numeber of samples.
preprocessDataForMCMC
MODEL	''	none	Model to fit.
F1	[]	none	Initial frequency for the analysis.
F2	[]	none	Final frequency for the analysis.
NUMERIC OUTPUT	0	0 1	Set to true to produce pure Matlab matrices as outputs
NAVS	5	none	The number of averages to use when calculating PSD and CPSD.
WIN	0	0 1	Windowing the data.
WIN TYPE	'BH92'	'Rectangular' 'Welch' 'Bartlett' 'Hanning' 'Hamming' 'Nuttall3' 'Nuttall4' 'Nuttall3a' 'Nuttall3b' 'Nuttall4a' 'Nuttall4b' 'Nuttall4c' 'BH92' 'SFT3F' 'SFT3M' 'FTNI' 'SFT4F' 'SFT5F' 'SFT4M' 'FTHP' 'HFT70' 'FTSRS' 'SFT5M' 'HFT90D' 'HFT95' 'HFT116D' 'HFT144D' 'HFT169D' 'HFT196D' 'HFT223D' 'HFT248D' 'Kaiser' 'levelledHanning'	Choose the type of the spectral window.
PSLL	150	none	Only if 'win' is set to 'true'. If you choose a 'kaiser' window, you can also specify the peak-sidelobe-level.
LEVELORDER	2	none	Only if 'win' is set to 'true'. If you choose a 'levelledHanning' window, you can also specify the order of the contraction.

Example
plist('METHOD', 'RJMCMC', 'NEFF', [[]], 'THETAMAP', [[]], 'SAMPLE', [true], 'LOG PARAMETERS', '', 'PCA', [false], 'M', [100], 'DISCARD', [0], 'INNAMES', '', 'OUTNAMES', '', 'MODELS', '', 'FITPARAMS', [[]], 'INPUT', '', 'N', [1000], 'COV', [0.0001], 'NOISE', '', 'MODELFREQDEPENDENT', [true], 'SEARCH', [true], 'FREQUENCIES', [[]], 'FSOUT', [[]], 'HEAT', [1], 'TC', '', 'X0', [[]], 'RANGE', [[]], 'BURNIN', [1], 'JUMPS', [[]], 'PLOT BF', [[]], 'PLOT PARAMETERS', [[]], 'DEBUG', [false], 'OUTMODEL', '', 'SCALE MATRIX', '', 'LOG-LIKELIHOOD', '', 'FREQUENCIES VECTOR', [200], 'INMODEL', '', 'FPRINT', [100], 'MODEL', '', 'F1', [[]], 'F2', [[]], 'NUMERIC OUTPUT', [false], 'NAVS', [5], 'WIN', [false], 'WIN TYPE', 'BH92', 'PSLL', [150], 'LEVELORDER', [2])

Example

plist('METHOD', 'RJMCMC', 'NEFF', [[]], 'THETAMAP', [[]], 'SAMPLE', [true], 'LOG PARAMETERS', '', 'PCA', [false], 'M', [100], 'DISCARD', [0], 'INNAMES', '', 'OUTNAMES', '', 'MODELS', '', 'FITPARAMS', [[]], 'INPUT', '', 'N', [1000], 'COV', [0.0001], 'NOISE', '', 'MODELFREQDEPENDENT', [true], 'SEARCH', [true], 'FREQUENCIES', [[]], 'FSOUT', [[]], 'HEAT', [1], 'TC', '', 'X0', [[]], 'RANGE', [[]], 'BURNIN', [1], 'JUMPS', [[]], 'PLOT BF', [[]], 'PLOT PARAMETERS', [[]], 'DEBUG', [false], 'OUTMODEL', '', 'SCALE MATRIX', '', 'LOG-LIKELIHOOD', '', 'FREQUENCIES VECTOR', [200], 'INMODEL', '', 'FPRINT', [100], 'MODEL', '', 'F1', [[]], 'F2', [[]], 'NUMERIC OUTPUT', [false], 'NAVS', [5], 'WIN', [false], 'WIN TYPE', 'BH92', 'PSLL', [150], 'LEVELORDER', [2])

Some information of the method matrix/modelSelect are listed below:
Class name	matrix
Method name	modelSelect
Category	Signal Processing
Package name	ltpda
VCS Version	967b0eec0dece803a81af8ef54ad2f8c784b20b2
Min input args	1
Max input args	-1
Min output args	1
Max output args	-1
Can be used as modifier	1
Supported numeric types	{'double'}

Method: matrix/mchNoisegenFilter

Method: matrix/polyfit

Method matrix/modelSelect

Parameter Description

Default

Example