reverted estimate_tau to maximize marginalized tau

Author: rsexton2

basicrta/gibbs.py

@@ -714,16 +714,17 @@ def estimate_tau(self):
         taus = 1 / rp.rates[rp.labels == index]
         wts = rp.weights[rp.labels == index]
         ci = confidence_interval(taus)
-        #h = np.histogram(taus, bins='sqrt')
-        #indmax = h[0].argmax()
-        #val = 0.5 * (h[1][:-1][indmax] + h[1][1:][indmax])
-
-        wbins = np.histogram_bin_edges(wts, bins=bintype)
-        rbins = np.histogram_bin_edges(taus, bins=bintype)
-        vals, ws, rs = np.histogram2d(wts, taus, bins=[wbins,rbins])
-        indmax = np.unravel_index(vals.argmax(), vals.shape)
-        rval = 0.5 * (rs[:-1] + rs[1:])[indmax[1]]
-        return [ci[0], rval, ci[1]]
+        h = np.histogram(taus, bins='sqrt')
+        indmax = h[0].argmax()
+        val = 0.5 * (h[1][:-1][indmax] + h[1][1:][indmax])
+        
+        # Used for finding maximum of weight vs tau 2d distribution
+        #wbins = np.histogram_bin_edges(wts, bins=bintype)
+        #rbins = np.histogram_bin_edges(taus, bins=bintype)
+        #vals, ws, rs = np.histogram2d(wts, taus, bins=[wbins,rbins])
+        #indmax = np.unravel_index(vals.argmax(), vals.shape)
+        #val = 0.5 * (rs[:-1] + rs[1:])[indmax[1]]
+        return [ci[0], val, ci[1]]
 
     def plot_surv(self, scale=1, remove_noise=False, save=False, xlim=None,
                   ylim=(1e-6, 5), xmajor=None, xminor=None):