Null hypothesis


Active Member
Hi Andreavdinu,

I will give you some hints:

You don't compare subject before taking the medicine and after, but a proportion of success (what is the distribution?)
You should compare the proportion of success of two groups, what are the groups?


Active Member
Generally, you have the two options you mention, both are valid, and you need to see which one is more appropriate to your experiment
I will start with words instead of exact math.

1. Two tails
H0: the medicine doesn't influence.
H1: the medicine do influence,
2. One Tail:
H0: the medicine doesn't influence
H1: the medicine reducing headaches

In option 1, H1 the medicine may also increase the headaches. But we are not interested in this option. we don't really care if the medicine doesn't reduce the headaches or only not influence, in both cases we won't take it.

H0: p<=0.5
H1: p>0.5

This is the right tail, and p-value is p(z>Z) = 1- P(z<Z)
you reject H0 if p-value<0.05

After you solve it you need to do a sanity check, for example, if p=0.8 instead of 0.6 is is more likely the medicine works, so you should get a smaller p-value