I have 2 hypotheses for the same thing. I'm basically testing two theories.

H1: there will be no difference between 2 groups - they will both do the same thing
H2: there will be a difference.

Now I'm testing this with a t-test.

My question is...(drumroll)........

Do I divide the P values by 2 (i.e. one-tailed) or leave them as they were (two-tailed)??????????????????????????????????????

Isn't this just one hypothesis, with you null hypothesis (H0) being that there is no difference? That said, what I know about stats, back of stamp, matchbox, etc etc...

normally I would say yes. But I'm doing this whole intro as a theory x vs theory y (in a battle to the death (from boredom) ) So I would need to put both hypotheses. This is all supervisors crazy idea, so don't blame me! :-)

I guess are you saying I should go with one-tailed, and then if it wasn't significant (which it is) then it would support the other hypothesis?

From what I know, your H1 is the null hypo. Your H1 should be H0 and H2 should be H1. So if there is no diff between the two groups then you confirm the null hypo, which should be H0. What stat books do you use. I recommend Andy field's 'bible', it sorts out lot of stat problems without sweats.;-)

oh my, is it two tailed?
whats the answer
love satchi

A two-tailed is the one that is normally always used to allow for the fact that either group 1 or group 2 has the higher mean. One-tailed tests are only used if your alternative hypothesis is that one group (i.e. you know which one) has a bigger mean, which is rarely justifiable. In whichever stats package you are using there should be an option for doing a one-tailed test instead of the default two-tailed.

I don't think there is an option in SPSS. Andy Field just says you divide the p value by 2 for a one-tailed.

I think I'm going to go with 1-tailed in my results and clearly mark that its one-tailed. and maybe state in my text the two-tailed (and clearly label it)

pls contact [email protected]

Thanks
Guidebuddha Team

Andy Fields is my hero :-) Not just for the content of his big bad SPSS book, but also because it is heavy enough for someone to be seriously hurt should I be tempted to throw it at them after the 100th failure of getting my stats right :-s

I always thought if you have a clear idea in which direction you expect the effext then you can use one tailed, but if you don't know whether group A wil be higher or group B (i.e just expect there to be a difference but don;t know in which direction) you'd use two tailed. That said you don't often see one tailed tests used. I alwsays feel a bit like it's cheating (even though I know it's not). I haven'y made a firm decision of what to do in mine....need to have a stats chat with supervisor.

Anyhoo if you do use one tailed you simply divide the p value by 2.

I'm going with one-tailed. The hypotheses are difficult, because I'm testing two theories, but I reckon one-tailed is the safest bet, I'm sure the reviewers will flag it if they think its an issue.

Hi Sneaks, I think you've found a solution already, but if not here's my advice. You're only testing one hypothesis, as one of your hypotheses you mention is known as the "null hypothesis", whilst the other is the one you expect to find support for. You should probably go with the two-tailed test because you don't hypotheses a direction in the differences, just that there will be a difference.

======= Date Modified 09 Oct 2010 09:32:00 =======
my supervisor re-edited my work AGAIN yesterday and has removed one hypothesis, so only testing one (which I find non-sig), so just moving all the literature about the other into the discussion (where it was originally, about 10 edits ago)

the problem before I think was that it wasn't really the null vs the real hypothesis. because H1 was "there will be no difference between groups, they will BOTH score 5", whereas hypothesis 2 was "there will be a difference, group 1 will score 0 and group 2 will score -4" - so although one said there will be no difference, it wasn't really the null, because its specifying the type of relationship too. confusing!

anyway problem solved, in a very annyoying way! :-s