def reset_max_error(itvc_entry, fit): sel = international_tables_stols <= fit.stol + 1.e-6 gaussian_fit = scitbx.math.gaussian.fit( international_tables_stols.select(sel), itvc_entry.table_y.select(sel), itvc_entry.table_sigmas.select(sel), fit) fit.max_error = flex.max(gaussian_fit.significant_relative_errors())
def run(gaussian_fit_pickle_file_names, itvc_file_name, kissel_dir): itvc_tab = None if (itvc_file_name is not None): itvc_tab = itvc_section61_io.read_table6111(itvc_file_name) fits = read_pickled_fits(gaussian_fit_pickle_file_names) #easy_pickle.dump("all_fits.pickle", fits) for k, v in fits.parameters.items(): print "# %s:" % k, v print max_errors = flex.double() labeled_fits = [] n_processed = 0 for label in expected_labels(kissel_dir): try: fit_group = fits.all[label] except Exception: print "# Warning: Missing scattering_type:", label else: print "scattering_type:", label prev_fit = None for fit in fit_group: if (prev_fit is not None): if (fit.stol > prev_fit.stol): print "# Warning: decreasing stol" elif (fit.stol == prev_fit.stol): if (fit.max_error < prev_fit.max_error): print "# Warning: same stol but previous has larger error" prev_fit = fit fit.sort().show() gaussian_fit = None if (itvc_tab is not None and label != "O2-"): entry = itvc_tab.entries[label] sel = international_tables_stols <= fit.stol + 1.e-6 gaussian_fit = scitbx.math.gaussian.fit( international_tables_stols.select(sel), entry.table_y.select(sel), entry.table_sigmas.select(sel), fit) elif (kissel_dir is not None): file_name = os.path.join( kissel_dir, "%02d_%s_rf" % (tiny_pse.table(label).atomic_number(), label)) tab = kissel_io.read_table(file_name) sel = tab.itvc_sampling_selection() & (tab.x <= fit.stol + 1.e-6) gaussian_fit = scitbx.math.gaussian.fit( tab.x.select(sel), tab.y.select(sel), tab.sigmas.select(sel), fit) if (gaussian_fit is not None): max_errors.append( flex.max(gaussian_fit.significant_relative_errors())) labeled_fits.append(labeled_fit(label, gaussian_fit)) n_processed += 1 print if (n_processed != len(fits.all)): print "# Warning: %d fits were not processed." % (len(fits.all) - n_processed) print if (max_errors.size() > 0): print "Summary:" perm = flex.sort_permutation(data=max_errors, reverse=True) max_errors = max_errors.select(perm) labeled_fits = flex.select(labeled_fits, perm) quick_summary = {} for me, lf in zip(max_errors, labeled_fits): print lf.label, "n_terms=%d max_error: %.4f" % ( lf.gaussian_fit.n_terms(), me) quick_summary[lf.label + "_" + str(lf.gaussian_fit.n_terms())] = me if (me > 0.01): fit = lf.gaussian_fit re = fit.significant_relative_errors() for s, y, a, r in zip(fit.table_x(), fit.table_y(), fit.fitted_values(), re): comment = "" if (r > 0.01): comment = " large error" print "%4.2f %7.4f %7.4f %7.4f %7.4f%s" % (s, y, a, a - y, r, comment) print print
def run(gaussian_fit_pickle_file_names, itvc_file_name, kissel_dir): itvc_tab = None if (itvc_file_name is not None): itvc_tab = itvc_section61_io.read_table6111(itvc_file_name) fits = read_pickled_fits(gaussian_fit_pickle_file_names) #easy_pickle.dump("all_fits.pickle", fits) for k,v in fits.parameters.items(): print "# %s:" % k, v print max_errors = flex.double() labeled_fits = [] n_processed = 0 for label in expected_labels(kissel_dir): try: fit_group = fits.all[label] except Exception: print "# Warning: Missing scattering_type:", label else: print "scattering_type:", label prev_fit = None for fit in fit_group: if (prev_fit is not None): if (fit.stol > prev_fit.stol): print "# Warning: decreasing stol" elif (fit.stol == prev_fit.stol): if (fit.max_error < prev_fit.max_error): print "# Warning: same stol but previous has larger error" prev_fit = fit fit.sort().show() gaussian_fit = None if (itvc_tab is not None and label != "O2-"): entry = itvc_tab.entries[label] sel = international_tables_stols <= fit.stol + 1.e-6 gaussian_fit = scitbx.math.gaussian.fit( international_tables_stols.select(sel), entry.table_y.select(sel), entry.table_sigmas.select(sel), fit) elif (kissel_dir is not None): file_name = os.path.join(kissel_dir, "%02d_%s_rf" % ( tiny_pse.table(label).atomic_number(), label)) tab = kissel_io.read_table(file_name) sel = tab.itvc_sampling_selection() & (tab.x <= fit.stol + 1.e-6) gaussian_fit = scitbx.math.gaussian.fit( tab.x.select(sel), tab.y.select(sel), tab.sigmas.select(sel), fit) if (gaussian_fit is not None): max_errors.append( flex.max(gaussian_fit.significant_relative_errors())) labeled_fits.append(labeled_fit(label, gaussian_fit)) n_processed += 1 print if (n_processed != len(fits.all)): print "# Warning: %d fits were not processed." % ( len(fits.all) - n_processed) print if (max_errors.size() > 0): print "Summary:" perm = flex.sort_permutation(data=max_errors, reverse=True) max_errors = max_errors.select(perm) labeled_fits = flex.select(labeled_fits, perm) quick_summary = {} for me,lf in zip(max_errors, labeled_fits): print lf.label, "n_terms=%d max_error: %.4f" % ( lf.gaussian_fit.n_terms(), me) quick_summary[lf.label + "_" + str(lf.gaussian_fit.n_terms())] = me if (me > 0.01): fit = lf.gaussian_fit re = fit.significant_relative_errors() for s,y,a,r in zip(fit.table_x(),fit.table_y(),fit.fitted_values(),re): comment = "" if (r > 0.01): comment = " large error" print "%4.2f %7.4f %7.4f %7.4f %7.4f%s" % (s,y,a,a-y,r,comment) print print