2
$\begingroup$

I ran Tassel3 and I filtered results with p-value not more than 0.05. This way, it is ok to draw a Manhattan plot. However, for a QQ-plot there is a problem.

Say I have 40 thousands SNPs, after filtering, about 1,800 SNPs were kept in the output file. Now how do I draw QQ-plot because of partial p-values?

ppoints only generate points within (0,1). I think the resulting points should be in the filtered range (0,0.05) with the same number (1,800) too. runif can do the trick, but it is random, each time it generates different results. I think this would affect the QQ-plot line.

I figured out three ways to solve it but I am not sure if it is right. Hope you guys can give some advices.

the following are three methods:

  1. 0.05*ppoint(1800)
  2. ppoints(1800*100/0.05, a=0.05), then I only get the former 1/20 results.
  3. ppoints(40k), and use those p-values below 0.05 in plotting.
$\endgroup$
3
$\begingroup$

You are right not to use random points. Version 3 will provide what you want. But as ppoints(40k) is just producing 40k equally-spaced values between 0 and 1, it's not too difficult to directly code what you want. To bypass the use of ppoints() altogether, use e.g. (-0.5+1:1800)/40000; this gives the expected position (under the null) of the 1800 smallest p-values, from a sample of 40000 p-values.

$\endgroup$
  • $\begingroup$ yes, thank you. a small trick, but a big difference. great! I prefer not using T3 to draw plot for I have more than one hundred traits' results to draw, and I haven't take my time to learn how to use command parameters of T3 to do this batchly. that's why I just use the results of T3 and use R to draw plots batchly. $\endgroup$ – eastasiax Feb 29 '12 at 1:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.