[pymvpa] RFE & Permutation

[Matthias Ekman]

Dear PyMVPA experts,

I tried to set up a RFE-Analysis with Permutationtest. But this results
in (implausible) low p-values. I guess there is smth. wrong with my code
(working with PyMVPA 0.4.3 and Linux)?

rfe_clf = LinearCSVMC(C=1)

rfesvm_split = SplitClassifier(rfe_clf)
clf = \
	FeatureSelectionClassifier(
	clf = LinearCSVMC(C=1),
	feature_selection = RFE(
	
sensitivity_analyzer=rfesvm_split.getSensitivityAnalyzer(combiner=FirstAxisMean,
transformer=N.abs),		
		transfer_error=ConfusionBasedError(
			rfesvm_split,
			confusion_state="confusion"),
			stopping_criterion=FixedErrorThresholdStopCrit(0.20),
		feature_selector=FractionTailSelector(
			0.2, mode='discard', tail='lower'),
		update_sensitivity=True))
		
splitter = NFoldSplitter(cvtype=1)		
NO_PERMUTATIONS=1000

cv = CrossValidatedTransferError(
		TransferError(clf),
		splitter, 	null_dist=MCNullDist(permutations=NO_PERMUTATIONS,tail='left'),
		enable_states=['confusion'])


Btw, how should this line look like for upcoming version 0.5.0?:
sensitivity_analyzer=rfesvm_split.getSensitivityAnalyzer(combiner=FirstAxisMean,
transformer=N.abs

I guess "transformer=N.abs" becomes "mapper=absolute_features()", but
what about the combiner? I am getting a TypeError here:

TypeError: Unexpected keyword argument combiner=<function FirstAxisMean
at 0x1929de8> for BoostedClassifierSensitivityAnalyzer: 4
states:states{base_sensitivities raw_results null_t null_prob+}. Valid
parameters are ['base_sensitivities', 'raw_results', 'null_t', 'null_prob']

Thanks in advance!
 Matthias