[R-sig-eco] quick stats question

[Scott Foster]

Dear Jonathan,

I think, without knowing details, that you may want to test that the 
variance between samples is what you would expect under a binomial with 
p=0.5 males.  After all, if the number of males is not binomial (since 
there is heterogeneity amongst units) then one would expect that the 
sample variance is higher than the binomial expected variance.  This is 
not 100% definitive as there are other reasons that data may be 
over-dispersed but, to me, it would be fairly conclusive.

The remaining issue is to work out how to perform the test of variance.  
I'm sure that there is some asymptotic test that could be used but I 
would prefer a permutation test: simulate lots of data sets according to 
a binomial p=0.5 *and your experimental/survey design*, calculate the 
variance for each one, and see where your observed variance lies with 
respect to this null distribution.

HTH,

Scott

On 12/02/11 01:13, Jonathan Hughes wrote:
>
> Dear all,
>
> This is both an R and a statistics question. I want to test whether males and females of a given species tend to co-occur in a given sampling unit more frequently than expected by chance. I'm thinking about using a binomial distribution with p as the sex ratio of the entire population. So, even though the population sex ratio is close to 50:50, each sampling unit would have disproportionately more males than females. Given that the test is not about p per se, how would you go about testing this hypothesis?
>
> Thanks!
>
> Jonathan
>   		 	   		
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-ecology mailing list
> R-sig-ecology at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
>

-- 
Scott Foster
CSIRO Mathematics, Informatics and Statistics
GPO Box 1538
Castray Esplanade
Hobart 7001
Tasmania
Australia

Phone:     (03) 6232 5178
Fax:       (03) 6232 5000
Email:     scott.foster at csiro.au