REALLY REALLY URGENT one tailed/two tailed help required

#1
Hi guys,

am in need of some very urgent help! My dissertation is due in on Monday and i have just come across something which means I have an awful lot of changes to make!

My hypothesis for my study is (i believe) one tailed;

Horses which win the Epsom Derby will have a significantly better stud career than horses which are placed in the Epsom Derby regardless of racing career performance.

However, i have never been taught or shown any examples using a one tailed hypothesis :( so i have been blissfully pottering along using spss to analyse my data and looking at the Asymp. Sig. (2-tailed) for my mann whitney U tests and Sig. (2-tailed) for my independent samples t-tests, because that was what the university tutorials said. HOWEVER i have just been flicking through discovering statistics using SPSS (Field 2005) and it says that if the hypothesis is one tailed, the sig. value in the t-test should be devided by 2 to give the one tailed significance. So I go back through the university tutorials for some advice but they are all two tailed examples :( problem that I'm having now is;

a) i might be reporting the wrong p values
b) as deviding my 2 tailed p values by 2 would leave them <0.05 this would change many of my results; making lots of relationships significant where they currently are not (which would be a good thing in my case!)
c) i might be looking like an utter idiot as i have been writing my hypothesis in all it's one tailed obviousness but reporting two tailed significances!

soo...

any advice? I cannot get in touch with my supervisor and my thesis is due on Monday! I am confident I could change all the data in the time but dont know whether I should :S realistically, I could change my hypothesis to;

There will be a significant difference between the stud careers of horses which win the Epsom Derby and horses which are placed in the Epsom Derby regardless of racing career performance.

but i dont know whether this would be acceptable or not :/ I read somewhere that many researchers suggest that one tailed hypotheses should only be used in unavoidable circumstance;

e.g. the height of 15 year old girls will be signifiantly greater than the height of 2 year old girls

so that confuses me even more as to whether I should change it!

Apologies if this makes little sense, I am incredibly rushed and stressed after this little discovery!

Any advice would be appreciated,

Elle
 

Dragan

Super Moderator
#2
Hi guys,

am in need of some very urgent help! My dissertation is due in on Monday and i have just come across something which means I have an awful lot of changes to make!

My hypothesis for my study is (i believe) one tailed;

Horses which win the Epsom Derby will have a significantly better stud career than horses which are placed in the Epsom Derby regardless of racing career performance.

However, i have never been taught or shown any examples using a one tailed hypothesis :( so i have been blissfully pottering along using spss to analyse my data and looking at the Asymp. Sig. (2-tailed) for my mann whitney U tests and Sig. (2-tailed) for my independent samples t-tests, because that was what the university tutorials said. HOWEVER i have just been flicking through discovering statistics using SPSS (Field 2005) and it says that if the hypothesis is one tailed, the sig. value in the t-test should be devided by 2 to give the one tailed significance. So I go back through the university tutorials for some advice but they are all two tailed examples :( problem that I'm having now is;

a) i might be reporting the wrong p values
b) as deviding my 2 tailed p values by 2 would leave them <0.05 this would change many of my results; making lots of relationships significant where they currently are not (which would be a good thing in my case!)
c) i might be looking like an utter idiot as i have been writing my hypothesis in all it's one tailed obviousness but reporting two tailed significances!

soo...

any advice? I cannot get in touch with my supervisor and my thesis is due on Monday! I am confident I could change all the data in the time but dont know whether I should :S realistically, I could change my hypothesis to;

There will be a significant difference between the stud careers of horses which win the Epsom Derby and horses which are placed in the Epsom Derby regardless of racing career performance.

but i dont know whether this would be acceptable or not :/ I read somewhere that many researchers suggest that one tailed hypotheses should only be used in unavoidable circumstance;

e.g. the height of 15 year old girls will be signifiantly greater than the height of 2 year old girls

so that confuses me even more as to whether I should change it!

Apologies if this makes little sense, I am incredibly rushed and stressed after this little discovery!

Any advice would be appreciated,

Elle
Elle, calm down - there is no reason to panic. That is, there's no reason to change anything. Leave the p-values as they are i.e. two-tailed.

If someone asks about a one-tail t-test then just tell them to take your p-value and cut it in half....good grief.
 
#3
Elle, calm down - there is no reason to panic. That is, there's no reason to change anything. Leave the p-values as they are i.e. two-tailed.

If someone asks about a one-tail t-test then just tell them to take your p-value and cut it in half....good grief.
ok...so perhaps I'm a little embarrassed now at my slight over reacting but it is a stressful time :$

Is there any chance that you could perhaps give an opinion as to whether you think it would be wise to change the wording of my hypothesis in my actual written thesis so that it reads more like a two tailed hypothesis? or do you think that this doesn't really matter?

Thanks so much for your quick reply, I'm panicking a fair bit less now :)

Thanks again,

Elle
 

TheEcologist

Global Moderator
#4
ok...so perhaps I'm a little embarrassed now at my slight over reacting but it is a stressful time :$

Is there any chance that you could perhaps give an opinion as to whether you think it would be wise to change the wording of my hypothesis in my actual written thesis so that it reads more like a two tailed hypothesis? or do you think that this doesn't really matter?

Thanks so much for your quick reply, I'm panicking a fair bit less now :)

Thanks again,

Elle

If I was your supervisor I wouldn't be all that worried. Though changing your hypothesis would be easy. Although changing the p-values should be simple enough as well, so why not?

Myself, I would opt for changing the p-values because (as Dragan says) its as simple as just 'cutting em in half'. You should evaluate if you then need to change your discussion/conclusions (which could be more work).

I guess it just boils down to what you think makes the best thesis.
 

TheEcologist

Global Moderator
#5
but i dont know whether this would be acceptable or not :/ I read somewhere that many researchers suggest that one tailed hypotheses should only be used in unavoidable circumstance;

Elle
Thats nonsense, they should be used in the appropriate cases (>, <), like yours (where you a priori define an 'appropriate hypothesis').
 
#6
If I was your supervisor I wouldn't be all that worried. Though changing your hypothesis would be easy. Although changing the p-values should be simple enough as well, so why not?

Myself, I would opt for changing the p-values because (as Dragan says) its as simple as just 'cutting em in half'. You should evaluate if you then need to change your discussion/conclusions (which could be more work).

I guess it just boils down to what you think makes the best thesis.
Thanks for your response :)

If im honest it would be great to cut the p-values in half as then I would end up with significant results from almost all of my tests, meaning that I can reject null hypothesis and my research would have worked exactly how I wanted it to.

I just wasn't sure whether this was a valid way of doing things, my uni isn't too hot on the whole stats thing if I'm honest so I think they'd just rather show us an easy way and hope we didn't need to do it any other way! (not that dividing some numbers by 2 is difficult!). So would you say that it would be acceptable to present all my p-values as 1 tailed and just halve them all from what SPSS has given me?

With regards to changing my discussion/conclusions etc. if I'm honest being able to conclude these results as siugnificant would be worth having to pull a few more all nighters haha!
 
#7
Thats nonsense, they should be used in the appropriate cases (>, <), like yours (where you a priori define an 'appropriate hypothesis').
oh and as for that, it was off an internet page I'd dug up from goodness knows where so not all that reputable I guess, thought it seemed a little extreme!