How to do a hypothesis test for the mean with known standard deviation (in Python, using SciPy)

Task

Let’s say we are measuring a variable over a population, and we know its standard deviation $σ$ is known, and assume that the variable is normally distributed. We take a sample, $x_{1}, x_{2}, x_{3}, \dots, x_{k}$ , and compute its mean $\bar{x}$ . We want to determine if the sample mean is significantly different from, greater than, or less than some hypothesized value, such as a hypothesized population mean. How do we formulate an appropriate hypothesis test?

Related tasks:

Solution

We will use the following fake data, but you can insert your actual data in its place. We have a sample of just 5 values and an assumed population standard deviation of 3.

sample = [31, 44, 28, 25, 40]  # sample data
pop_std = 3                    # population standard deviation

We also choose a value $0 \leq α \leq 1$ as our Type I error rate. We’ll let $α$ be 0.05 here, but you can change that in the code below.

Two-tailed test

In a two-tailed test, we compare the unknown population mean to a hypothesized value $m$ using the null hypothesis $H_{0} : μ = m$ . Here we’ll use $m = 30$ , but you can replace that value in the code below with the hypothesis relevant for your situation.

from scipy import stats
import numpy as np
m = 30                                         # hypothesized mean
n = len(sample)                                # number of observations
xbar = np.mean(sample)                         # sample mean
test_stat = (xbar - m) / (pop_std/np.sqrt(n))  # test statistic
2*stats.norm.sf(abs(test_stat))                # two-tailed p-value

0.007290358091535614

The $p$ -value, 0.00729, is less than $α$ , so we have evidence to reject the null hypothesis and conclude that the actual population mean $μ$ is significantly different from the hypothesized value of $m = 30$ .

Right-tailed test

In a right-tailed hypothesis test, the null hypothesis is that the population mean is greater than or equal to a chosen value, $H_{0} : μ \geq m$ .

Most of the code below is the same as above, but we repeat it to make it easy to copy and paste. Only the computation of the $p$ -value changes.

from scipy import stats
import numpy as np
m = 30                                         # hypothesized mean
n = len(sample)                                # number of observations
xbar = np.mean(sample)                         # sample mean
test_stat = (xbar - m) / (pop_std/np.sqrt(n))  # test statistic
stats.norm.sf(abs(test_stat))                  # right-tailed p-value

0.003645179045767807

The $p$ -value, 0.003645, is less than $α$ , so we have evidence to reject the null hypothesis and conclude that the actual population mean $μ$ is significantly less than the hypothesized value of $m = 30$ .

Left-tailed test

In a left-tailed hypothesis test, the null hypothesis is that the population mean is less than or equal to a chosen value, $H_{0} : μ \leq m$ .

Most of the code below is the same as above, but we repeat it to make it easy to copy and paste. Only the computation of the $p$ -value changes.

from scipy import stats
import numpy as np
m = 30                                         # hypothesized mean
n = len(sample)                                # number of observations
xbar = np.mean(sample)                         # sample mean
test_stat = (xbar - m) / (pop_std/np.sqrt(n))  # test statistic
stats.norm.sf(-abs(test_stat))                 # left-tailed p-value

0.9963548209542322

The $p$ -value, 0.99635, is greater than $α$ , so wwe do not have sufficient evidence to conclude that $μ > m$ and should continue to accept the null hypothesis, that $μ \leq m$ .

Content last modified on 24 July 2023.

See a problem? Tell us or edit the source.

Contributed by:

Elizabeth Czarniak (CZARNIA_ELIZ@bentley.edu)
Nathan Carter (ncarter@bentley.edu)