def resample(self, data=[], niter=5):
if len(data) == 0:
self.low = -sample_pareto(-self.x_m_low, self.alpha_low)
self.high = sample_pareto(self.x_m_high, self.alpha_high)
else:
for itr in range(niter):
# resample high, fixing low
self.x_m, self.alpha = self.x_m_high, self.alpha_high
super(Uniform, self).resample(data)
# tricky: flip data and resample 'high' again
self.x_m, self.alpha = -self.x_m_low, self.alpha_low
self.low, self.high = self.high, self.low
super(Uniform, self).resample(self._flip_data(data))
self.low, self.high = self.x_m_low - self.high, self.low
def resample(self,data=[],niter=5):
if len(data) == 0:
self.low = -sample_pareto(-self.x_m_low,self.alpha_low)
self.high = sample_pareto(self.x_m_high,self.alpha_high)
else:
for itr in xrange(niter):
# resample high, fixing low
self.x_m, self.alpha = self.x_m_high, self.alpha_high
super(Uniform,self).resample(data)
# tricky: flip data and resample 'high' again
self.x_m, self.alpha = -self.x_m_low, self.alpha_low
self.low, self.high = self.high, self.low
super(Uniform,self).resample(self._flip_data(data))
self.low, self.high = self.x_m_low - self.high, self.low
def resample(self, data=[]):
self.high = sample_pareto(*self._posterior_hypparams(
*self._get_statistics(data)))
return self
def resample(self,data=[]):
self.high = sample_pareto(
*self._posterior_hypparams(*self._get_statistics(data)))
return self