pim/entropic-spring
1
0
Fork 0
mirror of https://gitlab.com/pimnelissen/entropic-spring.git synced 2025-11-30 19:33:07 +01:00
Code Issues Packages Projects Releases Wiki Activity

update histogram style

Browse Source
This commit is contained in:
Pim Nelissen
2025-10-27 15:40:12 +01:00
parent 394a8eba78
commit ef95114fe8
1 changed files with 14 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
I.py
View File

@ -20,7 +20,8 @@ else:
lengths[i] = band.length/band.a
# histogram
values, bin_edges = np.histogram(lengths, bins=int(2*N), range=(-N, N))
bin_edges = np.arange(-N-1, N+1, 2)
values, _ = np.histogram(lengths, bins=bin_edges)
p_mc = values / values.sum() # normalise histogram
bin_centers = bin_edges[:-1] + np.diff(bin_edges)/2
@ -33,14 +34,21 @@ def P(N, L_over_a):
"""
return scp.special.binom(N, (N+L_over_a)/2) / 2**N
p_true = P(N, bin_centers-.5)
# Calculate p_true, scale with NUM_BANDS,
# mask low statistics bins,
# calculate chi^2/ndf
chi2 = np.sum((p_mc - p_true)**2 / p_true)
p_true = P(N, bin_centers)
p_true_scaled = p_true*np.sum(values)
values_masked = values[values > 5]
p_true_scaled_masked = p_true_scaled[values > 5]
chi2 = np.sum((values_masked - p_true_scaled_masked)**2 / p_true_scaled_masked)
chi2_over_ndf = chi2 / (len(values_masked)-1)
# plot hist vs true dist.
plt.plot(bin_centers-.5, p_true, label="$P(L)$")
plt.step(bin_centers, p_mc, color='r', alpha = 0.5, label="$\hat{P}(L)$")
plt.text(-45, 0.07, f"$\chi^2 = {round(chi2, 3)}$\n({len(lengths):.0E} samples)")
plt.plot(bin_centers, p_true, label="$P(L)$")
plt.bar(bin_centers, p_mc, color='r', alpha = 0.5, label="$\hat{P}(L)$")
plt.text(-45, 0.07, f"$\chi^2/ndf = {round(chi2_over_ndf, 3)}$\n({len(lengths):.0E} samples)")
plt.legend()
plt.xlabel("$L/a$")
plt.ylabel("P($L/a$)")
@ -53,7 +61,6 @@ plt.figure()
plt.scatter(bin_centers, p_ratio, marker='x')
plt.hlines(1, *xlim, color='r', linestyle='--')
plt.xlim(xlim)
plt.ylim(0.5, 1.5)
plt.xlabel("$L/a$")
plt.ylabel("$\hat{P}(L)/P(L)$")
plt.show()