Re: P values

[Guest guest]

HI Smita and friends,I think the current discussion is very well structured and because all of these fall on a continuum, we can keep the discussions about too high P-values for the stage when we discuss hypothesis testing. But, meanwhile, if there are thoughts that one wants to share, we can keep on going. Smita, if you want to get started on that, feel free.

/Arin

On Sun, Jul 11, 2010 at 4:48 AM, smita mali <smt_mali@...> wrote:

 

Hello sir,

Here are the answers from my side, any correction, suggestion or further explaination is welcome.

 

1.    What will be if we get p= 0.05 in any study?

This situation is like we toss the coin and instead of getting head or tail, coin stands on an edge. So what to do? Toss the coin again.

 

While doing clinical research what is more important is to justify the outcome clinically and support it statistically. If in any study p = 0.05 then I think clinical correlation is more important and accordingly the results could be compared with previous similar studies.

 

2.    As Arin proposed can we think of importance of 0.05, 0.04, 0.03, 0.02, 0.01 etc. separately?

Importance of p value 0.05, 0.04, 0.03, 0.02, 0.01 is that this makes significance level stricter and thus reduces the chances of false positive results. But this leads to increase in the sample size.

 

3.    Yes, and add to that, since we are asking that question, what happens if we get p values that are very high, say above 0.95, say 0.96, 0.97, and so on...?

4.    What's about fate of Ho and Ha when p = 0.05?

5.   Who first postulated the cut off of 0.05 and why?

 

 

I will keep the last three questions for our coming discussion on testing hypothesis which will be moderated by Dr Arin Basu. Is it ok sir?

 

Regards,

Dr Smita Mali