Why bother with Hypothesis Testing?
- Ashish Sharma
- 2 days ago
- 3 min read
In classrooms up and down the country, A level students sit and listen to teachers trying their best to explain hypothesis testing across the 3 types in the A-level Curriculum:
Binomial Hypothesis testing
Correlation Coefficient
Sample means for a Normal Distribution
Invariably students wonder what is the point of all this, the unclear terminology, the weird steps and non-committal conclusions at the end.
Well, what if I told you that Hypothesis testing plays a big role in the real world and actually helps various industries make important and strategic decisions?
Here are some examples where it can be used in practice:
1. Quality Control in Manufacturing
A factory produces metal rods that should be 10 cm long on average.A supervisor takes a random sample of 40 rods and measures their lengths.
Null hypothesis (H₀): The mean length is 10 cm.
Alternative hypothesis (H₁): The mean length is not 10 cm.
If the sample mean is too far from 10 cm (relative to the standard deviation), the factory concludes the machine is faulty and needs recalibration.
Why it matters: Prevents defective products, saves money, and keeps customers happy.
2. Medical Trials — Testing Whether a New Drug Works
A pharmaceutical company wants to know if a new painkiller is more effective than the existing one.
H₀: The new drug has the same average effect as the current one.
H₁: The new drug has a greater average effect.
They test the drug on a group of volunteers and compare results to the existing treatment using a hypothesis test.
Why it matters: Almost all medical treatments go through statistical hypothesis testing before approval.
3. Marketing — Does a New Advert Improve Sales?
A business launches a new advert and wants to know if it increases average daily sales.
H₀: The advert makes no difference to average sales.
H₁: The advert increases average sales.
They compare sales before and after the advert (or run an A/B test), and use a hypothesis test to see whether any differences are statistically significant.
Why it matters: Helps companies avoid spending money on ineffective campaigns.
4. Environmental Science — Is Air Pollution Increasing?
An environmental agency measures particulate matter (PM2.5) levels each year. They suspect pollution has risen since last year.
H₀: Pollution levels are unchanged.
H₁: Pollution levels have increased.
If the sample data shows a statistically significant rise, the agency may introduce policies to reduce emissions.
Why it matters: Hypothesis tests drive environmental policy and public health decisions.
5. Sports Analytics — Is a Player’s Performance Really Improving?
A coach wants to know if a football player’s shooting accuracy has genuinely improved or if recent success is just luck.
H₀: The player’s accuracy is unchanged.
H₁: The player’s accuracy has improved.
They compare proportions (a binomial hypothesis test) using past and current performance data.
Why it matters: Teams use statistics to decide on training, contracts, and tactics.
Ok- these may seem like theoretical examples. However I have also seen first hand how hypothesis testing is used in industry with the use of data science, python and statistics.
One example I have recently been analysing has been through a project looking at a Diabetes dataset and using a Pearson R hypothesis test.
This test is used to check whether there is a linear relationship between 2 quantitative variables. The null hypothesis is that there is NO correlation and the alternative is that there is (in either direction- 2 tailed test).
In particular my study focussed on whether the relationship between BMI and Glucose levels are statistically significant. The scatter graph shows a weak positive correlation which is supported by the PMCC value (r) of 0.233. I also calculated the P-value for this test and executed the project within Python as it is the tool of choice for statistical analysis (alongside R):
The P value is very small which set against even a very low significance level of 1% would be enough to reject the null hypothesis and show evidence to indicate a correlation between BMI and Glucose levels.
Although this research is not very high level study it does show that use cases for Hypothesis testing are diverse. Real business decisions can be taken off the back of such tests which impact people in many ways.
There is a vast array of literature that exists around hypothesis testing and its particular nuances (especially with P values).
So the next time you see a hypothesis test- remember this blog!
Comments