# [ Statistical Analysis with SciPy ]

Importing and Setup

● Import SciPy stats module: from scipy import stats

● Import NumPy for array operations: import numpy as np
● Set random seed for reproducibility: np.random.seed(42)

Descriptive Statistics

● Mean: np.mean(data)
● Median: np.median(data)
● Mode: stats.mode(data)
● Variance: np.var(data)
● Standard deviation: np.std(data)
● Range: np.ptp(data)
● Interquartile range: stats.iqr(data)
● Skewness: stats.skew(data)
● Kurtosis: stats.kurtosis(data)
● Coefficient of variation: stats.variation(data)
● Geometric mean: stats.gmean(data)
● Harmonic mean: stats.hmean(data)
● Trimmed mean: stats.trim_mean(data, 0.1)
● Percentile: np.percentile(data, 75)
● Quantile: np.quantile(data, [0.25, 0.5, 0.75])

Probability Distributions

● Normal distribution PDF: stats.norm.pdf(x, loc=0, scale=1)

● Normal distribution CDF: stats.norm.cdf(x, loc=0, scale=1)
● Normal distribution inverse CDF: stats.norm.ppf(q, loc=0, scale=1)
● Generate normal random numbers: stats.norm.rvs(loc=0, scale=1,
size=1000)
● Uniform distribution PDF: stats.uniform.pdf(x, loc=0, scale=1)
● Uniform distribution CDF: stats.uniform.cdf(x, loc=0, scale=1)
● Generate uniform random numbers: stats.uniform.rvs(loc=0, scale=1,
size=1000)
● Exponential distribution PDF: stats.expon.pdf(x, scale=1)
● Exponential distribution CDF: stats.expon.cdf(x, scale=1)

By: Waleed Mousa

 ● Generate exponential random numbers: stats.expon.rvs(scale=1,
size=1000)
● Poisson distribution PMF: stats.poisson.pmf(k, mu=1)
● Poisson distribution CDF: stats.poisson.cdf(k, mu=1)
● Generate Poisson random numbers: stats.poisson.rvs(mu=1, size=1000)
● Binomial distribution PMF: stats.binom.pmf(k, n, p)
● Binomial distribution CDF: stats.binom.cdf(k, n, p)
● Generate binomial random numbers: stats.binom.rvs(n, p, size=1000)
● Chi-square distribution PDF: stats.chi2.pdf(x, df)
● Chi-square distribution CDF: stats.chi2.cdf(x, df)
● Generate chi-square random numbers: stats.chi2.rvs(df, size=1000)
● Student's t-distribution PDF: stats.t.pdf(x, df)
● Student's t-distribution CDF: stats.t.cdf(x, df)
● Generate Student's t random numbers: stats.t.rvs(df, size=1000)
● F-distribution PDF: stats.f.pdf(x, dfn, dfd)
● F-distribution CDF: stats.f.cdf(x, dfn, dfd)
● Generate F random numbers: stats.f.rvs(dfn, dfd, size=1000)

Hypothesis Testing

● One-sample t-test: stats.ttest_1samp(data, popmean)

● Independent two-sample t-test: stats.ttest_ind(data1, data2)
● Paired t-test: stats.ttest_rel(data1, data2)
● One-way ANOVA: stats.f_oneway(data1, data2, data3)
● Two-way ANOVA: stats.f_oneway(*(group for name, group in
data.groupby(['factor1', 'factor2'])))
● Chi-square goodness of fit test: stats.chisquare(observed, expected)
● Chi-square test of independence:
stats.chi2_contingency(contingency_table)
● Shapiro-Wilk test for normality: stats.shapiro(data)
● Anderson-Darling test for normality: stats.anderson(data)
● Kolmogorov-Smirnov test: stats.kstest(data, 'norm')
● Mann-Whitney U test: stats.mannwhitneyu(data1, data2)
● Wilcoxon signed-rank test: stats.wilcoxon(data1, data2)
● Kruskal-Wallis H-test: stats.kruskal(data1, data2, data3)
● Friedman test: stats.friedmanchisquare(data1, data2, data3)
● Levene's test for equality of variances: stats.levene(data1, data2)
● Bartlett's test for equality of variances: stats.bartlett(data1, data2)
● Fligner-Killeen test for equality of variances: stats.fligner(data1,
data2)

By: Waleed Mousa

Correlation and Regression

● Pearson correlation coefficient: stats.pearsonr(x, y)

● Spearman rank correlation: stats.spearmanr(x, y)
● Kendall's tau: stats.kendalltau(x, y)
● Simple linear regression: stats.linregress(x, y)
● Multiple linear regression: stats.linregress(X, y)
● Polynomial regression: np.polyfit(x, y, deg=2)
● R-squared (coefficient of determination): 1 - (np.sum((y - y_pred)**2)
/ np.sum((y - np.mean(y))**2))
● Adjusted R-squared: 1 - ((1 - r_squared) * (n - 1) / (n - k - 1))
● F-statistic: ((r_squared / (k - 1)) / ((1 - r_squared) / (n - k)))
● Durbin-Watson statistic: stats.durbin_watson(residuals)

Non-parametric Methods

● Kernel density estimation: stats.gaussian_kde(data)

● Bootstrap sample: stats.bootstrap((data,), np.mean, n_resamples=1000)
● Jackknife resampling: stats.jackknife(data, np.mean)
● Permutation test: stats.permutation_test((data1, data2),
stats.ttest_ind)

Multivariate Analysis

● Principal Component Analysis: from sklearn.decomposition import PCA;

PCA().fit_transform(X)
● Canonical correlation analysis: from sklearn.cross_decomposition
import CCA; CCA().fit(X, Y).transform(X, Y)
● MANOVA: from statsmodels.multivariate.manova import MANOVA;
MANOVA.from_formula('y1 + y2 ~ group', data=data).mv_test()
● Hotelling's T-squared test: stats.hotelling_t2(X1, X2)

Time Series Analysis

● Autocorrelation: stats.autocorr(data)
● Partial autocorrelation: from statsmodels.tsa.stattools import pacf;
pacf(data)
● Augmented Dickey-Fuller test: from statsmodels.tsa.stattools import
adfuller; adfuller(data)
● KPSS test: from statsmodels.tsa.stattools import kpss; kpss(data)

By: Waleed Mousa

 ● Granger causality test: from statsmodels.tsa.stattools import
grangercausalitytests; grangercausalitytests(data, maxlag=5)

Bayesian Statistics

● Bayes factor: stats.bayes_mvs(data)

● Bayesian Information Criterion (BIC): stats.bic(residuals)
● Akaike Information Criterion (AIC): stats.aic(residuals)

Sampling and Experimental Design

● Simple random sample: np.random.choice(population, size=n,

replace=False)
● Stratified sample: from sklearn.model_selection import
StratifiedShuffleSplit; StratifiedShuffleSplit(n_splits=1,
test_size=0.3).split(X, y)
● Cluster sample: from sklearn.cluster import KMeans;
KMeans(n_clusters=k).fit_predict(X)
● Systematic sample: population[::k]
● Latin square design: stats.latin_square(n)

Power Analysis

● Power of t-test: stats.ttest_ind_solve_power(effect_size=0.5,

nobs1=100, alpha=0.05, ratio=1.0, alternative='two-sided')
● Power of ANOVA: stats.f_oneway_solve_power(dfnum=2, dfden=27,
alpha=0.05, effect_size=0.25)
● Sample size calculation for t-test:
stats.ttest_ind_solve_power(effect_size=0.5, power=0.8, alpha=0.05,
ratio=1.0, alternative='two-sided')

Reliability Analysis

● Cronbach's alpha: from statsmodels.stats.inter_rater import

fleiss_kappa; fleiss_kappa(data)
● Intraclass correlation coefficient: stats.ttest_ind(group1, group2)

Effect Size Calculations

By: Waleed Mousa

 ● Cohen's d: (np.mean(group1) - np.mean(group2)) /
np.sqrt((np.std(group1, ddof=1)**2 + np.std(group2, ddof=1)**2) /
2)
● Eta-squared: ss_effect / (ss_effect + ss_error)
● Odds ratio: (a * d) / (b * c)
● Risk ratio: (a / (a + b)) / (c / (c + d))

Data Transformation

● Z-score normalization: stats.zscore(data)

● Min-max scaling: (data - np.min(data)) / (np.max(data) -
np.min(data))
● Box-Cox transformation: stats.boxcox(data)
● Yeo-Johnson transformation: stats.yeojohnson(data)
● Logarithmic transformation: np.log1p(data)

Outlier Detection

● Z-score method: np.abs(stats.zscore(data)) > 3

● Interquartile range (IQR) method: (data < Q1 - 1.5 * IQR) | (data > Q3
+ 1.5 * IQR)
● Modified Z-score method: 0.6745 * (data - np.median(data)) /
stats.median_abs_deviation(data) > 3.5
● Grubbs' test: stats.grubbs(data)

Confidence Intervals

● Normal distribution CI: stats.norm.interval(alpha=0.95,

loc=np.mean(data), scale=stats.sem(data))
● T-distribution CI: stats.t.interval(alpha=0.95, df=len(data)-1,
loc=np.mean(data), scale=stats.sem(data))
● Binomial proportion CI: stats.binom.interval(n=len(data),
p=np.mean(data), alpha=0.05)
● Poisson CI: stats.poisson.interval(alpha=0.95, mu=np.mean(data))

Survival Analysis

● Kaplan-Meier estimator: from lifelines import KaplanMeierFitter;

KaplanMeierFitter().fit(durations, event_observed)
● Cox proportional hazards model: from lifelines import CoxPHFitter;
CoxPHFitter().fit(df, duration_col='T', event_col='E')

By: Waleed Mousa

 ● Log-rank test: from lifelines.statistics import logrank_test;
logrank_test(durations_1, durations_2, event_observed_1,
event_observed_2)

Spatial Statistics

● Moran's I: from pysal.explore import esda; esda.Moran(y, w).I

● Geary's C: from pysal.explore import esda; esda.Geary(y, w).C
● Getis-Ord G: from pysal.explore import esda; esda.G(y, w).G

Multivariate Normality Tests

● Mardia's test: from statsmodels.stats.multivariate_normal import

mardia; mardia(data)
● Henze-Zirkler test: from statsmodels.stats.multivariate_normal import
henze_zirkler; henze_zirkler(data)

Robust Statistics

● Median absolute deviation: stats.median_abs_deviation(data)

● Huber's M-estimator: from statsmodels.robust import scale;
scale.huber(data)
● Theil-Sen estimator: from scipy.stats import theilslopes;
theilslopes(y, x)

Factor Analysis

● Exploratory Factor Analysis: from factor_analyzer import

FactorAnalyzer; FactorAnalyzer().fit(data)
● Confirmatory Factor Analysis: from statsmodels.stats.factor import
FactorAnalysis; FactorAnalysis().fit(data)

Cluster Analysis

● K-means clustering: from sklearn.cluster import KMeans;

KMeans(n_clusters=k).fit(X)
● Hierarchical clustering: from scipy.cluster.hierarchy import linkage;
linkage(X, method='ward')
● DBSCAN clustering: from sklearn.cluster import DBSCAN;
DBSCAN().fit(X)

By: Waleed Mousa

Time Series Decomposition

● Seasonal decomposition: from statsmodels.tsa.seasonal import

seasonal_decompose; seasonal_decompose(data, model='additive')

Statistical Process Control

● Control chart (X-bar chart): from statsmodels.stats.stattools import

control_chart; control_chart(data)

Meta-Analysis

● Fixed effects meta-analysis: from statsmodels.stats.meta_analysis

import CombineResults; CombineResults.combine_effects(effects,
variances)
● Random effects meta-analysis: from statsmodels.stats.meta_analysis
import CombineResults; CombineResults.combine_effects(effects,
variances, method='random')

Structural Equation Modeling

● Path analysis: from statsmodels.stats.sem import SEM;

SEM.from_formula('y ~ x1 + x2', data=data).fit()

Item Response Theory

● 1PL (Rasch) model: from psychometrics import irt;

irt.twopl(difficulty, discrimination=1, ability)
● 2PL model: from psychometrics import irt; irt.twopl(difficulty,
discrimination, ability)

Multilevel Modeling

● Random intercept model: from

statsmodels.regression.mixed_linear_model import MixedLM;
MixedLM.from_formula('y ~ x', groups='group', data=data).fit()

Statistical Quality Control

● Capability analysis: from statsmodels.stats.stattools import
cpk_index; cpk_index(data, lower=lsl, upper=usl)
● Process capability index: (usl - lsl) / (6 * np.std(data, ddof=1))
By: Waleed Mousa
Nonlinear Regression

● Curve fitting: from scipy.optimize import curve_fit;

curve_fit(lambda x, a, b: a * np.exp(b * x), x_data, y_data)

Statistical Tests for Circular Data

● Rayleigh test: from scipy.stats import rayleigh; rayleigh.fit(data)

● Watson's U2 test: from astropy.stats import watson_u2;
watson_u2(data)

Extreme Value Analysis

● Generalized extreme value distribution fit: from scipy.stats import

genextreme; genextreme.fit(data)
● Peak over threshold analysis: from scipy.stats import genpareto;
genpareto.fit(data[data > threshold])

Functional Data Analysis

● Functional principal component analysis: from skfda.decomposition

import FPCA; FPCA().fit_transform(data)

Statistical Learning Theory

● Support Vector Machine: from sklearn.svm import SVC; SVC().fit(X, y)

● Cross-validation: from sklearn.model_selection import
cross_val_score; cross_val_score(model, X, y, cv=5)

Copulas

● Gaussian copula: from scipy.stats import multivariate_normal;

multivariate_normal.cdf(data)
● Clayton copula: from copulas.multivariate import
GaussianMultivariate;
GaussianMultivariate().fit(data).probability_density(data)

Stochastic Processes

● Brownian motion simulation: np.cumsum(np.random.normal(0, 1,

size=1000))

By: Waleed Mousa

 ● Ornstein-Uhlenbeck process: from scipy.integrate import odeint;
odeint(lambda y, t, theta, mu, sigma: theta * (mu - y), y0, t,
args=(theta, mu, sigma))

Causal Inference

● Propensity score matching: from sklearn.linear_model import

LogisticRegression; LogisticRegression().fit(X,
treatment).predict_proba(X)[:, 1]
● Difference-in-differences estimation: np.mean(post_treatment -
pre_treatment) - np.mean(post_control - pre_control)

Spatial Point Pattern Analysis

● Ripley's K function: from astropy.stats import RipleysKEstimator;

RipleysKEstimator(area=area).evaluate(data)

Statistical Network Analysis

● Erdős-Rényi random graph model: from networkx.generators.random_graphs

import erdos_renyi_graph; erdos_renyi_graph(n, p)

By: Waleed Mousa