Streamer272/state-soc-cross
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

🔨 Adds kruskal and reworks chi square

Browse Source
This commit is contained in:
Daniel Svitan 2025-05-21 20:09:56 +02:00
parent 48b1ac4753
commit 5f6656d710
1 changed files with 29 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

71
analysis.py
View File

@ -76,31 +76,13 @@ with open("dataset.txt") as stream:
# 2-7 - first to last place county idxs
data_original = np.array(raw_data)
# table where counties are rows and counts of placements are columnes
# #1 | #2 | ...
# BA | 5 | 4 | ...
# ZA | 9 | 8 | ...
# KE | 4 | 6 | ...
# as a row-first 2d numpy array (first dimension will represent counties, second counts of placements)
# data = np.zeros((counties_c, 5)) # 5 because top five
# for sample in data_original:
# results = sample[2:7]
# for placement_idx, county_idx in enumerate(results):
# data[county_idx, placement_idx] += 1
# data is table where rows represent placement and columns county index
# 1st | 5 | 1 | 2 | ...
# 2nd | 3 | 0 | 7 | ...
# 3rd ...
# data = np.zeros((5, years * categories_c)) # same as (5, len(data_original))
# for i, sample in enumerate(data_original):
# results = sample[2:7]
# for j in range(5):
# data[j][i] = results[j]
# ------
# H0: each county wins proportionally to its population
# H1: some counties win more than others after adjusting for population
# wins per county
# goodness-of-fit problem using Chi Square
# goodness-of-fit problem using chi square
# based on observed vs expected frequency
observed = np.zeros(counties_c)
for sample in data_original:
results = sample[2:7]
@ -110,34 +92,39 @@ for sample in data_original:
print("Observed before adjusting for population:")
print(observed)
# micro-wins per capita (because wins would be a tiny number)
# micro-wins per capita (because wins per capita would be a tiny number)
for i in range(len(observed)):
observed[i] = observed[i] / counties_population[i] * 1_000_000
observed[i] = observed[i] / counties_population[i] * 1_000_000 # `*million` because unit is micro
print("Observed after adjusting for population:")
print("Observed after adjusting for population [micro-wins per capita]:")
print(observed)
expected = np.ones_like(observed) * (sum(observed) / len(observed))
print("Expected after adjusting for population:")
print("Expected:")
print(expected)
chi2, p = stats.chisquare(f_obs=observed, f_exp=expected)
print(f"Chi-square = {chi2:.2f}, p-value = {p:.4f}")
print(f"Chi-Square = {chi2:.4f}, p-value = {p:.8f}")
print("")
# H0: county and placement are independent
# H1: county and placement are not independent
# ------
# H0: there is no difference between county placements (once they do get placed)
# H1: some counties get better placements than others
# kruskal wallis
# places grouped by counties
# print("\nAttempting Chi-Square test")
# chi2, p, dof, expected = stats.chi2_contingency(data)
# data is table where rows represent county and columns placements
# BA | 5 | 1 | 4 | ...
# TN | 2 | 2 | 3 | ...
# TT ...
data = []
for _ in range(8):
data.append([])
for sample in data_original:
results = sample[2:7]
for i, res in enumerate(results):
data[res].append(i + 1) # range is [1 - 5]
# print(f"Chi-Square Statistic: {chi2}")
# print(f"p-value: {p}")
# print(f"Degrees of Freedom: {dof}")
# print("Expected Frequencies:\n", expected)
# print("\nAttempting Fisher's Exact test")
# oddsratio, p_value = stats.fisher_exact(data)
# print(f"Odds Ratio: {oddsratio}")
# print(f"p-value: {p_value}")
F, p = stats.kruskal(*data)
print(f"Kruskal-Wallis F = {F:.4f}, p-value = {p:.8f}")