def exp1ModelBuild(dm):
"""
desc:
Creates optimal models for exp 3. A full interactive model for the
full dataset, and non-interactive models for each of the subsets,
split by scene type and instruction.
arguments:
dm:
type: DataMatrix.
"""
assert(exp in ['exp1', 'expAll'])
modelRandomEffects = '(1|file)'
modelCandidateFixedEffects = stats.modelCandidateFixedEffects[:-1]
lfe = stats.modelBuild(dm, suffix='.full',
modelRandomEffects=modelRandomEffects,
modelCandidateFixedEffects=modelCandidateFixedEffects)
model = 'salFrom ~ %s + pupilSize + (1+pupilSize|file)' % (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp1/model.indirect.csv')
model = 'salFrom ~ pupilSize + (1+pupilSize|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp1/model.direct.csv')
def exp1BestPupil(dm):
"""
desc:
Determines which of the various pupil-size transforms is the best
predictor of saliency.
arguments:
dm:
type: DataMatrix.
"""
modelRandomEffects = '(1|file)'
modelCandidateFixedEffects = stats.modelCandidateFixedEffects[:-1]
lfe = stats.modelBuild(dm, suffix='.full',
modelRandomEffects=modelRandomEffects,
modelCandidateFixedEffects=modelCandidateFixedEffects)
model = 'salFrom ~ %s + %%s + (1+%%s|file)' % (' + '.join(lfe))
stats.R.load(dm)
for m in range(1, 7):
_model = model % ('pupilSize%d' % m, 'pupilSize%d' % m)
lm = stats.R.lmer(_model, lmerVar='model%d' % m)
lm._print('model%d' % m)
lm.save('output/%s/bestPupil/lmer.pupilSize%d.csv' % (exp, m))
m1 = 1
for m2 in range(2, 7):
am = stats.R.anova('model%s' % m1, 'model%s' % m2)
am._print('%s - %s' % (m1, m2))
am.save('output/%s/bestPupil/anova.pupilSize%d-%d.csv' \
% (exp, m1, m2))
def exp2ModelBuild(dm):
"""
desc:
Creates optimal models for exp 2. A full interactive model for the
full dataset, and non-interactive models for each of the subsets,
split by scene type and instruction.
arguments:
dm:
type: DataMatrix.
"""
assert(exp == 'exp2')
modelRandomEffects = '(1|file)'
modelCandidateFixedEffects = stats.modelCandidateFixedEffects[:-1]
lfe = stats.modelBuild(dm, suffix='.full',
modelRandomEffects=modelRandomEffects,
modelCandidateFixedEffects=modelCandidateFixedEffects)
model = 'salFrom ~ %s + pupilSize*inst*sceneType + (1+pupilSize+inst+sceneType|file)' \
% (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp2/model.indirect.interactionMainRandom.csv')
model = 'salFrom ~ pupilSize*inst*sceneType + (1+pupilSize+inst+sceneType|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp2/model.direct.interactionMainRandom.csv')
model = 'salFrom ~ %s + pupilSize*inst + pupilSize*sceneType + (1+pupilSize+inst+sceneType|file)' \
% (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp2/model.indirect.splitInteractionMainRandom.csv')
model = 'salFrom ~ pupilSize*inst + pupilSize*sceneType + (1+pupilSize+inst+sceneType|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp2/model.direct.splitInteractionMainRandom.csv')
model = 'salFrom ~ %s + pupilSize + (1+pupilSize|file)' % (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp2/model.indirect.csv')
model = 'salFrom ~ pupilSize + (1+pupilSize|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp2/model.direct.csv')
for sceneType in dm.unique('sceneType'):
for inst in dm.unique('inst'):
_dm = dm.select('sceneType == "%s"' % sceneType) \
.select('inst == "%s"' % inst)
lfe = stats.modelBuild(_dm, suffix='.%s.%s' % (sceneType, inst),
modelRandomEffects=modelRandomEffects,
modelCandidateFixedEffects=modelCandidateFixedEffects)
model = 'salFrom ~ %s + pupilSize + (1+pupilSize|file)' \
% (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp2/model.indirect.%s.%s.csv' % (sceneType, inst))
model = 'salFrom ~ pupilSize + (1+pupilSize|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp2/model.direct.%s.%s.csv' % (sceneType, inst))
for sceneType in dm.unique('sceneType'):
_dm = dm.select('sceneType == "%s"' % sceneType)
lfe = stats.modelBuild(_dm, suffix='.%s' % sceneType,
modelRandomEffects=modelRandomEffects,
modelCandidateFixedEffects=modelCandidateFixedEffects)
model = 'salFrom ~ %s + pupilSize + (1+pupilSize|file)' \
% (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp2/model.indirect.%s.csv' % sceneType)
model = 'salFrom ~ pupilSize + (1+pupilSize|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp2/model.direct.%s.csv' % sceneType)
for inst in dm.unique('inst'):
_dm = dm.select('inst == "%s"' % inst)
lfe = stats.modelBuild(_dm, suffix='.%s' % inst,
modelRandomEffects=modelRandomEffects,
modelCandidateFixedEffects=modelCandidateFixedEffects)
model = 'salFrom ~ %s + pupilSize + (1+pupilSize|file)' \
% (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp2/model.indirect.%s.csv' % inst)
model = 'salFrom ~ pupilSize + (1+pupilSize|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp2/model.direct.%s.csv' % inst)
def exp3ModelBuild(dm):
"""
desc:
Creates optimal models for exp 3. A full interactive model for the
full dataset, and non-interactive models for each of the subsets,
split by scene type and instruction.
arguments:
dm:
type: DataMatrix.
"""
assert(exp == 'exp3')
modelRandomEffects = '(1|file)'
modelCandidateFixedEffects = stats.modelCandidateFixedEffects[:-1]
lfe = stats.modelBuild(dm, suffix='.full',
modelRandomEffects=modelRandomEffects,
modelCandidateFixedEffects=modelCandidateFixedEffects)
model = 'salFrom ~ %s + cond + (1+cond|file)' % (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp3/model.cond.indirect.csv')
model = 'salFrom ~ cond + (1+cond|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp3/model.cond.direct.csv')
model = 'salFrom ~ %s + pupilSize*cond + (1+pupilSize+cond|file)' \
% (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp3/model.indirect.interactionMainRandom.csv')
model = 'salFrom ~ pupilSize*cond + (1+pupilSize+cond|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp3/model.direct.interactionMainRandom.csv')
model = 'salFrom ~ %s + pupilSize + (1+pupilSize|file)' % (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp3/model.indirect.csv')
model = 'salFrom ~ pupilSize + (1+pupilSize|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp3/model.direct.csv')
for cond in dm.unique('cond'):
_dm = dm.select('cond == "%s"' % cond)
lfe = stats.modelBuild(_dm, suffix='.%s' % cond,
modelRandomEffects=modelRandomEffects,
modelCandidateFixedEffects=modelCandidateFixedEffects)
model = 'salFrom ~ %s + pupilSize + (1+pupilSize|file)' \
% (' + '.join(lfe))
lm = stats.R.lmer(model)
lm._print('Indirect model')
lm.save('output/exp3/model.indirect.%s.csv' % cond)
model = 'salFrom ~ pupilSize + (1+pupilSize|file)'
lm = stats.R.lmer(model)
lm._print('Direct model')
lm.save('output/exp3/model.direct.%s.csv' % cond)