A statistical test provides a mechanism for making quantitative decisions about a process or processes. It is a way to evaluate the evidence the data provides against a hypothesis.
I have created a flow chart that shows which statistical test to use depending on your data and test requirements. In this post, I will focus on how to perform these tests in Python. A statistical test examines two opposing hypotheses about a population: the null hypothesis and the alternative hypothesis. The null hypothesis states that a population parameter (such as the mean, the standard deviation, and so on) is equal to a hypothesized value. The alternative hypothesis states that a population parameter is smaller, greater, or different than the hypothesized value in the null hypothesis.
One-Sample Z test
One sample Z test is used to compare the population mean to a sample. Population standard deviation must be known to perform the Z test and data should be normally distributed.
x = sample mean
mu = population mean
s = population standard deviation
n = sample size
z = (x - mu)/(s / np.sqrt(n))
p = 1 - stats.norm.cdf(z)
if p < alpha **# Reject Null Hypothesis**
if p > alpha **# Failed to reject Null Hypothesis**
One-Sample T-test
One sample T-test is used to compare the population mean to a sample. This test is similar to the Z test as it checks the same thing. The difference between these two tests is that the Z test needs population standard deviation but the T-test needs sample standard deviation. If the sample size is greater than 30 One sample T-test is always used regardless of the known population standard deviation.
from scipy import stats
t_stat, pval = stats.ttest_1samp(sample, popmean)
critical_value = stats.t.ppf(1 - 0.05, df)
If t_stat > critical_value Reject Null Hypothesis
p **=** 1 **-** stats.t.cdf(t_stat, df)
If p < alpha **# Reject Null Hypothesis**
One-Sample Z test for Proportion
One-Sample Z test for proportion is used to compare an observed proportion to a theoretical one. The test of proportion can only be performed on categorical data.
p = proportion in sample
p0 = theoretical proportion
z_stat = (p - p0) / np.sqrt(p0 * (1 - p0) / n)
p = 1 - stats.norm.cdf(z)
if p < alpha **# Reject Null Hypothesis**
if p > alpha **# Failed to reject Null Hypothesis**
Two sample Z test for Proportion
Two Sample Z test for proportion is used to compare two proportions with a theoretical proportion.
p1 = first proportion in sample
p2 = second proportion in sample
p0 = theoretical proportion
z_stat = ((p1 - p2) - 0) / np.sqrt(p(1-p)((1/n1) + (1/n2)))
p = 1 - stats.norm.cdf(z)
if p < alpha **# Reject Null Hypothesis**
if p > alpha **# Failed to reject Null Hypothesis**
Two Samples T-Test
Two sample T-test is used to compare two samples. There are two types of Two-sample T-test, a Paired T-test, and a Pooled T-test.
Paired T-test
A paired T-test is used when both samples have different variances.
t_stat, pval = stats.ttest_ind(sample1, sample2, equal_var=False)
critical_value = stats.t.ppf(1 - 0.05, df)
If t_stat > critical_value Reject Null Hypothesis
p **=** 1 **-** stats.t.cdf(t_stat, df)
If p < alpha **# Reject Null Hypothesis**
Pooled T-test
The pooled T-test is used when both samples have equal or almost equal variances.
t_stat, pval = stats.ttest_ind(sample1, sample2)
critical_value = stats.t.ppf(1 - 0.05, df)
If t_stat > critical_value Reject Null Hypothesis
p **=** 1 **-** stats.t.cdf(t_stat, df)
If p < alpha **# Reject Null Hypothesis**
Two sample Z test
Two sample Z test is used to compare the means of two populations.
x1 = sample1 mean
x2 = sample2 mean
n1 = sample1 size
n2 = sample2 size
sigma1 = standard deviation of sample1
sigma2 = standard deviation of sample2
delta = hypothesized difference between the population means (0 if testing for equal means)
mu = population mean
s = population standard deviation
n = sample size
z_stat = (x1 - x2 - delta)/np.sqrt((sigma1**2/n1) + (sigma2**2/n2))
p = 1 - stats.norm.cdf(z_stat)
if p < alpha **# Reject Null Hypothesis**
if p > alpha **#** **Failed to reject Null Hypothesis**
Chi-Squared Test
The chi-squared test is used to compare the relationship between the two categorical (nominal) variables in a contingency table. There are three kinds of chi-squared tests: the test of independence, the goodness of fit, and the test of homogeneity. This test is applied to a contingency table of values in the dataset.
from scipy.stats import chi2_contingency
table = np.array([[16, 18, 16], [32, 24, 16]]).T #contingency table
stat, p, dof, expected = chi2_contingency(table)
critical_value = chi2.ppf(0.95, dof)
if stat >= critical_value **# Reject Null Hypothesis**
if p <= alpha **# Reject Null Hypothesis**
ANOVA
ANOVA test is used to compare means between three or more variables. It is the same as Independent T-test for two samples.
from statsmodels.formula.api import ols
import statsmodels.api as sm
anova = ols('var1~var2', data=data).fit()
anova_table = sm.stats.anova_lm(anova, type=2)
print(anova_table)
Using the above code and printing anova_table will give you an ANOVA table to interpret.
df sum_sq mean_sq F PR(>F)
var2 3.0 9.5059e+08 3.1686e+08 128.769622 **6.720391e-67**
Residual 727.0 1.7889e+09 2.4607e+06 NaN NaN
p_value = **6.720391e-67
**p_value <<<<< alpha
Reject Null Hypothesis
Conclusion
A statistical test is one of the key skills of Data Science. It can be very challenging to choose a statistical test for your data. I struggled most at figuring when to use which statistical test. When to use which statistical test depends on the nature of data(continuous or categorical, dependent or independent, paired or not paired).
Check this link for more information on statistical tests with Python.
Thank you for reading.
