These sets of controls are used to provide default values in this package.
Objects contain several parameters for methods.
Wei-Chen Chen and Ranjan Maitra.
The elements of .FC.CT are default values for main controls of
MixfMRI including
| Elements | Default | Usage |
algorithm | "apecma" | implemented algorithm |
optim.method | "BFGS" | optimization method |
model.X | "I" | cov matrix structure |
ignore.X | FALSE | if using voxel information |
check.X.unit | TRUE | if checking X in [0, 1] |
CONTROL | a list | see CONTROL next for details |
INIT | a list | see INIT next for details |
LRT | a list | see LRT next for details |
MPI.gbd | FALSE | if MPI speedup available |
common.gbd | TRUE | if X in common gbd format |
The elements of CONTROL are default values for optimization controls
of implemented EM algorithm including
| Elements | Default | Usage |
max.iter | 1000 | maximum number of EM iterations |
abs.err | 1e-4 | absolute error of convergence |
rel.err | 1e-6 | relative error of convergence |
debug | 1 | debugging level |
RndEM.iter | 10 | RndEM iterations |
exp.min | log(.Machine$double.xmin) | minimum exponential power |
exp.max | log(.Machine$double.xmax) | maximum exponential power |
sigma.ill | 1e-6 | ill condition limit |
DS.max | 1e+4 | maximum chol() cov matrix |
DS.min | 1e-6 | minimum chol() cov matrix |
The elements of INIT are default values or limitations for initial
parameters implemented for EM algorithm including
| Elements | Default | Usage |
min.1st.prop | 0.8 | minimum proportion of 1st cluster |
max.PV | 0.1 | maximum p-value for initialization |
BETA.alpha.min | 0 + 1e-6 | minimum value of alpha parameter of Beta distribution |
BETA.alpha.max | 1 - 1e-6 | maximum value of alpha parameter of Beta distribution |
BETA.beta.min | 1 + 1e-6 | minimum value of beta parameter of Beta distribution |
BETA.beta.max | 1e+6 | maximum value of beta parameter of Beta distribution |
max.try.iter | 10 | maximum retry iterations if result is unstable |
class.method | "prob.extned" | classification method at initializations |
The elements of LRT are default values or limitations for likelihood
ratio tests including
| Elements | Default | Usage |
H0.alpha | 1 | null hypothesis alpha parameter of Beta distribution |
H0.beta | 1 | null hypothesis beta parameter of Beta distribution |
H0.mean | 0.05 | null hypothesis mean of Beta distribution |
Chen, W.-C. and Maitra, R. (2021) “A Practical Model-based Segmentation Approach for Accurate Activation Detection in Single-Subject functional Magnetic Resonance Imaging Studies”, arXiv:2102.03639.
set.global(), fclust().