Advertised Sequences and View Counts

[Kirbyderp]

Ah, I see what you're referring to. Let me explain a few things.
First of all, you actually can do it with 2 samples of size 15. You could even do it with size 10.
You see, the standard deviation for the sampling distribution of x̄A-x̄N needs to be approximately Normal in order to do inference about a population. Now, the larger the sample sizes are, the more this helps, as the sampling distribution of the means will approach the Normal model as sample size increases due to the Central Limit Theorem. However, even if you have a small sample, if the distribution of the sample data are unimodal and roughly symmetric, then the sampling distribution will still be Normal.

I also want to explain how sample size directly relates to the variability of the sampling distribution. For the difference in means, it is , where σ represents standard deviation and n represents sample size. As you can probably see, the large sample size decreases the variation, as we're dividing by n. With less variation, a smaller difference could be detected.

All in all, while a larger sample size helps, it is not necessary.

Also, if you tl;dr this I will completely ignore the rest of your arguments.