<aside> 🔥 This is an exam point!
</aside>
<aside> 💡 This is a good to know!
</aside>
Collect, Analyse, Communicate
We want to see the effect of some treatment. There is thus a need compare the treatment effect using a treatment group and a control group. This eliminates the effect of time. This is called a controlled experiment.
<aside> 💡 Effect of Time: Sometimes, it may be tempting to use historical controls. However, there are a lot of confounders.
</aside>
If control and treatment groups are similar, a difference in response can be attributed to the treatment, else the treatment effect may be confounded with some other factor.
But how do we get similar groups? Often, for controlled experiments, subjects who refuse treatment or agree to treatment do so for a reason. There are thus confounding factors. Thus, only subjects who agree should be enrolled and assigned to control or treatment. And the process requires some assignment that makes these two groups as similar as possible.
This leads us to Randomised Assignment.
Using an impartial procedure based on chance, such as random draws without replacement, we can form two groups from the willing subjects randomly.