sage-bionetworks / bfrm Goto Github PK

Bayesian Factor Regression Model (BFRM) techniques from West group at Duke - wrappers written to integrate as an R package to call platform-specific executable.

C 2.52% C++ 73.24% Shell 22.16% R 2.09%

bfrm's Introduction

README DOCUMENT FOR THE bfrm PACKAGE

AUTHOR: BRIAN M. BOT

Bayesian Factor Regression Modelling (bfrm) - wrappers written to integrate bfrm algorithm from West group as an R package to call platform-specific executables.

bfrm's People

Contributors

Stargazers

Watchers

Forkers

erichhuang nagyistoce rosefu79

bfrm's Issues

add parseBFRM (or some better name) function?

# parseBFRM.R
# Functions to parse BFRM output into a list object of Factors by feature indices
# that meet a 0.99 posterior probability threshold

parseBFRM <- function(bfrmResult){
  print('Loading Posterior Probabilities')
  mPostPib <- bfrmResult@results$mPostPib
  numFacs <- dim(mPostPib)[2]
  facParse <- function(x){
    incFeatureLogical <- x > 0.99
    incFeatureInd <- grep('TRUE', incFeatureLogical)
    return(incFeatureInd)
  }
  print('Creating list of Factors by Features with Posterior Probability > 0.99')
  facList <- apply(mPostPib[ , 2:numFacs], 2, facParse)
  return(facList)
}

Error in checkSlotAssignment(object, name, value) : "evolmaximumvariablesperfactor" is not a slot in class “bfrmParam”

fooFactor <- evolve(egfrMat,
init = as.numeric(topProbeInd),
varThreshold = 0.85,
facThreshold = 0.95,
maxVarIter = 30,
minFacVars = 10,
maxFacVars = length(topProbeInd),
maxFacs = 50,
maxVars = length(topProbeInd)
)

evolve() needs to specify EvolVarInFile to work properly

I think I've pinned down where the bug is:
When "evol = 1", just as with the bfrm() function, the file of the "seed" indices has to be specified, in this case with the "evolvarinfile" argument. So the temporary scratch parameter file needs this line added in and a temporary scratch text file of these indices needs to be created.

minFacVars isn't a slot

Error in checkSlotAssignment(object, name, value) :
"minFacVars" is not a slot in class “bfrmParam”

document parameter defaults in .rd files

enabling transparency for all parameter values that are being set

Default values for 'priorpsia' and 'priorpsib' should be changed

Values should be:
priorpsia = 2
priorpsib = 0.005

convenience plotting functions

heatmap, others

"evolmaximumperfactors" is not a slot in class "bfrmParam"

fooFactor <- evolve(egfrMat,
                init = as.numeric(topProbeInd),
                maxVarIter = 30,
                maxFacs = 5,
                maxVars = length(topProbeInd))
Error in checkSlotAssignment(object, name, value) :
"evolmaximumperfactors" is not a slot in class “bfrmParam”

maxVars syntax error (extra parenthesis)

                maxVars = length(topProbeInd))
Error: unexpected ')' in " maxVars = length(topProbeInd))"