Why is the P-value of Fisher's exact test (Fisher's P) greater than one ?

[Mark]

Hi, anyone

 

Some P-values of Fisher's exact test (Fisher's P in the table batch_1.fst_0-1.tsv) are greater than one. I think the P-value should be within the interval 0 ~ 1. So, What's wrong with me? Are there some mistakes in my work?

I divided 17 samples into two groups. One of groups has 8 samples and the other 9 samples. After some work, the POPULATIONS program was used to compare the two groups as followings:

 

populations -t 36  -b 1 -k -p 10 -r 0.75 -f p_value \

           -P /home/ma/stacks/denovo_map_population_92_b1/   \

           -M /home/ma/stacks/populationmap.txt   

 

Near 0.3% of Fisher's Ps are greater than 1. 

I am also meeting this problem at other situations while using the POPULATIONS software. So I am wondering how I choose the statistical significant alleles between different populations. Could someone offer me some advice?

 

Mark

[Julian Catchen]

Hi Mark,

I have rechecked the implementation of Fisher's Exact Test in the
populations program and found that it is correct. Its calculations match
the implementation in R. The small number of values that are above 1
should be considered as a p-value of 1.0.

Best,

julian

Mark wrote:
> Hi, anyone
>
> Some P-values of Fisher's exact test (Fisher's P in the table
> batch_1.fst_0-1.tsv) are greater than one. I think the P-value should be
> within the interval 0 ~ 1. So, What's wrong with me? Are there some
> mistakes in my work?
>
> I divided 17 samples into two groups. One of groups has 8 samples and
> the other 9 samples. After some work, the POPULATIONS program was used
> to compare the two groups as followings:
>

> populations -t 36-b 1 -k -p 10 -r 0.75 -f p_value \

>
> -P /home/ma/stacks/denovo_map_population_92_b1/\
>
> -M /home/ma/stacks/populationmap.txt
>
> Near 0.3% of Fisher's Ps are greater than 1.
>

> <https://lh6.googleusercontent.com/-GENeU01eDN4/U7NRBa-ZmWI/AAAAAAAAAAo/BliqDQB6Ubw/s1600/P_distribution.png>

[Mark]

Dear Julian,
Thank you very much.

Best wishes.

 

Mark