1
$\begingroup$

I am doing a ab test. I want to test if a treatment is statistically better than a control. I want to split 30 subjects randomly and evenly into two groups. My way is first generate random number for these 30 subjects, and then sort them in an ascending order. Finally, assign first 15 subjects into A group, and the rest to B group. My questions are

  1. Are there any problems with this random assignment method ?
  2. Can I still use statistical test, like t test and chi square test under this setting ?
  3. Why is it rarely see people use this method ? Compare to this method, what are the benefits of simple random assignment ? (In web application setting)
$\endgroup$
1
  • $\begingroup$ There is nothing wrong with your procedure so long as your (pseudo-)random number generator (almost) never produces duplicate values $\endgroup$
    – Henry
    Nov 3, 2022 at 9:48

1 Answer 1

1
$\begingroup$

Your procedure is the same as if you had a bowl with two kinds of tickets: one kind would have a “$\le 15$” note and another kind “$> 15$”, $15$ of each kind, where your participants would randomly draw the tickets from the bowl. It is just another way of operationalizing the procedure of randomly assigning participants to equally sized groups.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.