Hi
I analzysed 15 populations and 10 microsat loci with Arlequin ver 3. I have already got observed and expected heterozygosity and corresponding P-value.
How can I perform Bonferroni correction for those data? However, there was no option to fix P-value in Arlequin soft. for like 0.05/0.1 before analyzing data. So, which P-value i need to select for Bonferroni correction?
Another question is that
in some cases, the value for observed heterozygosity was greater than expected heterozygosity in some loci. Is that affect my result and how?
Please, give some explanation.
I have attached the HWE file here. Please explain the bonferroni correction method with my example data.
Thanks
Rakeb
Why have you chosen Bonferroni's correction? Holm's method dominants it (and there are several other options). Assuming Bonferroni's is what you want, in R just do:
p.adjust(your.p.values, method="bonferroni")
where your.p.values is a vector of your p values. For example,
p.adjust(c(.005,.009,.05,.1,.2,.3),method="bonferroni")
produces
[1] 0.030 0.054 0.300 0.600 1.000 1.000
And for comparison
p.adjust(c(.005,.009,.05,.1,.2,.3),method="holm")
produces
[1] 0.030 0.045 0.200 0.300 0.400 0.400
There are lots of other corrections that can be used. This is a big topic with the number of tests often being very large. Efron's book on the topic is very good.
Hi Daniel
Thank you very much for your help. From my data I got the result as follows after running both.
p.adjust(c(0.14057,0.04561,0.0309,0.02863,0.07182,0.00066,0.21414,0.02691,0.00006,0.06448,0.0002),method="bonferroni")
[1] 1.00000 0.50171 0.33990 0.31493 0.79002 0.00726 1.00000
[8] 0.29601 0.00066 0.70928 0.00220
> p.adjust(c(0.14057,0.04561,0.0309,0.02863,0.07182,0.00066,0.21414,0.02691,0.00006,0.06448,0.0002),method="holm")
[1] 0.28114 0.22805 0.21528 0.21528 0.25792 0.00594 0.28114
[8] 0.21528 0.00066 0.25792 0.00200
However, my P-value for this data was 0.05. So, How do I know that which is significant? Is that the value which is less than 0.005?
Please let me know
Thanks
Rakeb
the common approach would be to treat the .00594, .0066, and .002 as "statistically significant". However, it is always trickier interpreting hypothesis testing when there is more than one hypothesis test.
You can also try qvalue Package in R. You can go through this link
https://bioconductor.org/packages/release/bioc/html/qvalue.html
There does not need correction! Сorrection need when we have repeated (multiple) comparisons, but it is just repeated measurements.
Dear Ed, I am curious about your distinction regarding comparisons and measurements. The corrections listed refer to dealing with a vector of probabilities. While there are some special considerations depending on how they are hypothesized to arise (e.g., if one is a control group), things like Bonferroni's inequality is a property of probabilities, so not specific to your distinction. Maybe I am just confused. Can you please clarify to help me to understand.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts