inferential-statistics

Inferential test selection and execution including t-test, ANOVA, chi-square, and non-parametric alternatives

الانتقال إلى التثبيت

المستودع: sencersoylu/scholar-flow
آخر نشاط في المصدر: ٢٠ مارس ٢٠٢٦ في ٠٧:٠٦
لغة SKILL.md المكتشفة: الإنجليزية
النجوم: ١
التفرعات: ٠

خيارات التثبيت

يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.

التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.

يتجاوز الأمر المباشر Prompt المخصّص للمراجعة. افحص المصدر قبل تشغيله.

npx skills add https://github.com/sencersoylu/scholar-flow --skill inferential-statistics

يبقى الأمر في سطر واحد. مرّر أفقيًا لمراجعته كاملًا قبل النسخ.

تفضّل نسخة محلية؟ نزّل الملفات المتاحة حاليًا لدى SkillsMP.

مراجعة ملفات المصدر

اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.

مستكشف الملفات

7 ملفات

SKILL.md

تعليمات المصدر · معاينة للقراءة فقط

العودة إلى البحث العودة إلى التثبيت

from scipy import stats
import pandas as pd
import numpy as np

# Example: comparing outcome between two groups
group1 = df[df['group'] == 'treatment']['outcome']
group2 = df[df['group'] == 'control']['outcome']

# Normality (per group)
for name, data in [('Treatment', group1), ('Control', group2)]:
    stat, p = stats.shapiro(data.dropna())
    print(f"{name}: Shapiro-Wilk W={stat:.4f}, p={p:.4f}")

# Homogeneity of variance
stat, p = stats.levene(group1.dropna(), group2.dropna())
print(f"Levene's test: F={stat:.4f}, p={p:.4f}")

from scipy import stats

# Standard t-test (assumes equal variances)
t_stat, p_value = stats.ttest_ind(group1, group2)

# Welch's t-test (does not assume equal variances -- generally preferred)
t_stat, p_value = stats.ttest_ind(group1, group2, equal_var=False)

print(f"t = {t_stat:.3f}, p = {p_value:.4f}")

# Effect size: Cohen's d
def cohens_d(g1, g2):
    n1, n2 = len(g1), len(g2)
    var1, var2 = g1.var(), g2.var()
    pooled_std = np.sqrt(((n1-1)*var1 + (n2-1)*var2) / (n1+n2-2))
    return (g1.mean() - g2.mean()) / pooled_std

d = cohens_d(group1.dropna(), group2.dropna())
print(f"Cohen's d = {d:.3f}")

# Non-parametric alternative to independent t-test
u_stat, p_value = stats.mannwhitneyu(group1, group2, alternative='two-sided')
print(f"U = {u_stat:.1f}, p = {p_value:.4f}")

# Effect size: rank-biserial correlation
n1, n2 = len(group1), len(group2)
r_rb = 1 - (2 * u_stat) / (n1 * n2)
print(f"Rank-biserial r = {r_rb:.3f}")

# Paired data (e.g., pre-post measurements)
t_stat, p_value = stats.ttest_rel(pre_scores, post_scores)
print(f"t = {t_stat:.3f}, p = {p_value:.4f}")

# Effect size: Cohen's d for paired data
diff = post_scores - pre_scores
d_paired = diff.mean() / diff.std()
print(f"Cohen's d (paired) = {d_paired:.3f}")

# Non-parametric alternative to paired t-test
stat, p_value = stats.wilcoxon(pre_scores, post_scores)
print(f"W = {stat:.1f}, p = {p_value:.4f}")

# Effect size: r = Z / sqrt(N)
n = len(pre_scores)
z = stats.norm.ppf(p_value / 2)
r = abs(z) / np.sqrt(n)
print(f"Effect size r = {r:.3f}")

# Parametric: one-way ANOVA
group_a = df[df['group'] == 'A']['outcome']
group_b = df[df['group'] == 'B']['outcome']
group_c = df[df['group'] == 'C']['outcome']

f_stat, p_value = stats.f_oneway(group_a, group_b, group_c)
print(f"F = {f_stat:.3f}, p = {p_value:.4f}")

# Effect size: eta-squared
import statsmodels.api as sm
from statsmodels.formula.api import ols

model = ols('outcome ~ C(group)', data=df).fit()
anova_table = sm.stats.anova_lm(model, typ=2)
ss_between = anova_table['sum_sq']['C(group)']
ss_total = anova_table['sum_sq'].sum()
eta_sq = ss_between / ss_total
print(f"Eta-squared = {eta_sq:.3f}")

# Post-hoc pairwise comparisons (Tukey HSD)
from statsmodels.stats.multicomp import pairwise_tukeyhsd

tukey = pairwise_tukeyhsd(df['outcome'], df['group'], alpha=0.05)
print(tukey)

from scipy import stats

# Welch's ANOVA (does not assume equal variances)
stat, p_value = stats.alexandergovern(group_a, group_b, group_c)
print(f"Welch's ANOVA: statistic={stat:.3f}, p={p_value:.4f}")

# Post-hoc: Games-Howell (for unequal variances)
# Use pingouin library
import pingouin as pg
posthoc = pg.pairwise_gameshowell(data=df, dv='outcome', between='group')
print(posthoc)

# Non-parametric alternative to one-way ANOVA
h_stat, p_value = stats.kruskal(group_a, group_b, group_c)
print(f"H = {h_stat:.3f}, p = {p_value:.4f}")

# Effect size: epsilon-squared
n = len(df)
k = df['group'].nunique()
epsilon_sq = (h_stat - k + 1) / (n - k)
print(f"Epsilon-squared = {epsilon_sq:.3f}")

# Post-hoc: Dunn's test with Bonferroni correction
from scikit_posthocs import posthoc_dunn
dunn = posthoc_dunn(df, val_col='outcome', group_col='group', p_adjust='bonferroni')
print(dunn)

# Contingency table
contingency = pd.crosstab(df['exposure'], df['outcome'])
print(contingency)

# Chi-square test
chi2, p_value, dof, expected = stats.chi2_contingency(contingency)
print(f"Chi-square = {chi2:.3f}, df = {dof}, p = {p_value:.4f}")

# Check expected counts
print(f"Minimum expected count: {expected.min():.1f}")
if expected.min() < 5:
    print("WARNING: Expected counts < 5. Consider Fisher's exact test.")

# Effect size: Cramer's V
n = contingency.sum().sum()
min_dim = min(contingency.shape) - 1
cramers_v = np.sqrt(chi2 / (n * min_dim))
print(f"Cramer's V = {cramers_v:.3f}")

# For 2x2 tables with small expected counts
contingency_2x2 = pd.crosstab(df['exposure'], df['outcome'])
odds_ratio, p_value = stats.fisher_exact(contingency_2x2)
print(f"Odds Ratio = {odds_ratio:.3f}, p = {p_value:.4f}")

# Pearson correlation (parametric)
r, p_value = stats.pearsonr(df['var1'], df['var2'])
print(f"Pearson r = {r:.3f}, p = {p_value:.4f}")

# Spearman correlation (non-parametric)
rho, p_value = stats.spearmanr(df['var1'], df['var2'])
print(f"Spearman rho = {rho:.3f}, p = {p_value:.4f}")

Method	Description	When to Use
Bonferroni	Divide alpha by number of tests	Conservative; few comparisons
Holm (step-down)	Sequential Bonferroni	Less conservative than Bonferroni; general use
Benjamini-Hochberg (FDR)	Controls false discovery rate	Many comparisons; exploratory analyses
Tukey HSD	For all pairwise ANOVA comparisons	Post-hoc after significant ANOVA
Dunnett	Compare all groups to one control	Multiple treatment groups vs. one control

from statsmodels.stats.multitest import multipletests

# Example: correcting multiple p-values
p_values = [0.01, 0.03, 0.04, 0.15, 0.22]

# Bonferroni
reject_bonf, pvals_bonf, _, _ = multipletests(p_values, method='bonferroni')

# Holm
reject_holm, pvals_holm, _, _ = multipletests(p_values, method='holm')

# Benjamini-Hochberg (FDR)
reject_fdr, pvals_fdr, _, _ = multipletests(p_values, method='fdr_bh')

for i, p in enumerate(p_values):
    print(f"p={p:.3f} | Bonf={pvals_bonf[i]:.4f} ({reject_bonf[i]}) | "
          f"Holm={pvals_holm[i]:.4f} ({reject_holm[i]}) | "
          f"FDR={pvals_fdr[i]:.4f} ({reject_fdr[i]})")

Groups	Related?	Parametric	Non-Parametric
2	Independent	Independent t-test	Mann-Whitney U
2	Paired/matched	Paired t-test	Wilcoxon signed-rank
>2	Independent	One-way ANOVA	Kruskal-Wallis
>2	Repeated measures	Repeated measures ANOVA

Groups	Related?	Parametric	Non-Parametric
2	Independent	Independent t-test	Mann-Whitney U
2	Paired/matched	Paired t-test	Wilcoxon signed-rank
>2	Independent	One-way ANOVA	Kruskal-Wallis
>2	Repeated measures	Repeated measures ANOVA

Expected counts	Groups	Test
All >= 5	2+	Chi-square test of independence
Any < 5, 2x2 table	2	Fisher's exact test
Any < 5, larger table	2+	Fisher-Freeman-Halton exact test
Paired/matched	2	McNemar's test

Data type	Test
Both continuous, normal	Pearson correlation (r)
Ordinal or non-normal	Spearman rank correlation (rho)
Both ordinal, small n	Kendall's tau

Effect Size	Small	Medium	Large
Cohen's d	0.2	0.5	0.8
Pearson r	0.1	0.3	0.5
Eta-squared	0.01	0.06	0.14
Cramer's V (df=1)	0.1	0.3	0.5
Cramer's V (df=2)	0.07	0.21	0.35
Cramer's V (df=3)	0.06	0.17	0.29
Odds Ratio	1.5	2.5	4.3

inferential-statistics

خيارات التثبيت

مراجعة ملفات المصدر