hypothesis testing - how do you know when to reject null hypothesis

Say the significance level is 5%

the probability you get in the binomial distribution is 0.056

0.056 > 0.05 so do you reject the null hypothesis or accept it... how do you know?

I just dont get it

Original post by m.s124

Say the significance level is 5%

the probability you get in the binomial distribution is 0.056

0.056 > 0.05 so do you reject the null hypothesis or accept it... how do you know?

I just dont get it

Never say 'accept'.

You either reject the null, or not.

If the p-value (which is what you calculated) is below the significance level, then you reject the null hypothesis. Otherwise, nothing happens.

I would recommend you look at some resources if your understanding on this topic is a bit shaky.

Reply 2

o_reo

OP

17

Original post by RDKGames

Never say 'accept'.

You either reject the null, or not.

If the p-value (which is what you calculated) is below the significance level, then you reject the null hypothesis. Otherwise, nothing happens.

I would recommend you look at some resources if your understanding on this topic is a bit shaky.

Thank you I will keep that in mind

if its below the significance level why do you reject it?
Everywhere ive looked it doesnt explain why

Original post by m.s124

Thank you I will keep that in mind

if its below the significance level why do you reject it?
Everywhere ive looked it doesnt explain why

The whole hypothesis testing idea is that the null hypothesis is a well-established statement about the population parameter of some distribution.

If you go ahead and sample enough points from the distribution, and calculate the same parameter for your sample, then you might think that if it's wildly different, then the population parameter must not be what it is currently.

So you ask: how likely is it, that under the assumption that the null hypothesis is true, that you ended up with your sample parameter being what it is?

If this probability is extremely low, i.e. below the level of significance, then you conclude that it is extremely unlikely that you ended up with some awkward samples due to luck. Instead, it suggests that there has been a shift by the population parameter itself away from what it currently thought to be.

(edited 3 years ago)

Reply 4

o_reo

OP

17

Original post by RDKGames

The whole hypothesis testing idea is that the null hypothesis is a well-established statement about the population parameter of some distribution.

If you go ahead and sample enough points from the distribution, and calculate the same parameter for your sample, then you might think that if it's wildly different, then the population parameter must not be what it is currently.

So you ask: how likely is it, that under the assumption that the null hypothesis is true, that you ended up with your sample parameter being what it is?

If this probability is extremely low, i.e. below the level of significance, then you conclude that it is extremely unlikely that you ended up with some awkward samples due to luck. Instead, it suggests that there has been a shift by the population parameter itself away from what it currently thought to be.