[GenABEL-dev] Bug #1247; ProbABEL: what to do with Rsq in info file

[L.C. Karssen]

Dear all,

In order to close this bug I've decided to go with option (1): fix the
documentation. This fix is committed in SVN r.971.


Best,

Lennart.



On 16-07-12 15:44, L.C. Karssen wrote:
> On 07/14/2012 04:13 PM, Yurii Aulchenko wrote:
>> I would go for (1) - when Rsq is indeed zero or very close to zero,
>> nonsense results are obtained. Generates a lot of confusion, and this
>> is why the check is there.
> 
> At the moment the test is:
> 
>  	if (fabs(mli.Rsq[csnp])<1.e-16) poly=0;
> 
> Isn't that threshold of 1e-16 too low? I observe in our imputations
> using the 1kG phase 1 alpha reference that only for monomorphic SNPs we
> get an Rsq of 0.0, the lowest Rsq for a non-monomorphic SNP is 1e-5.
> Wouldn't a cut-off of 0.3 (as is regularly done in meta-analysis) be
> better? Or, to give some room, use 0.1?
> 
> Anyway, don't you think we should print a warning message to stderr that
> says something like: "Warning: one or more of your SNPs have a very low
> imputation quality (Rsq < 0.1) and consequently their effect estimates
> are set to 'nan'"?
> 
> 
> Thanks,
> 
> Lennart.
> 
> 
>>
>> Yurii
>>
>> ----------------------
>> Yurii Aulchenko
>> (sent from mobile device)
>>
>> On 13 Jul 2012, at 19:46, "L.C. Karssen" <lennart at karssen.org> wrote:
>>
>>> Dear list,
>>>
>>> As pointed out on the forum
>>> (http://forum.genabel.org/viewtopic.php?f=10&t=663&p=1145#p1145) the
>>> ProbABEL manual states that the contents of the info file is not
>>> relevant for the ProbABEL output and is simply copied over.
>>>
>>> It turns out that this is not the case (see also bug #2147:
>>> https://r-forge.r-project.org/tracker/index.php?func=detail&aid=2147&group_id=505&atid=2058).
>>>
>>> In main.cpp line 751 we test for the Rsq values taken from the info
>>> file. The test is used to decide whether to output betas, se_betas etc.
>>> as 'nan' in the output files (see lines 771 and 836 in main.cpp).
>>>
>>> I see two possible solutions:
>>> 1) update the manual and say that the Rsq column in the info file should
>>> contain non-zero values. How many people will be using ProbABEL with
>>> 'fake' info files?
>>>
>>> 2) remove of modify the Rsq test in main.cpp. Suggestions are welcome on
>>> a sane way to implement this. Print a warning message?
>>>
>>>
>>> Thanks for thinking along,
>>>
>>> Lennart.
>>> --
>>> -----------------------------------------------------------------
>>> L.C. Karssen
>>> Utrecht
>>> The Netherlands
>>>
>>> lennart at karssen.org
>>> http://blog.karssen.org
>>>
>>> Stuur mij aub geen Word of Powerpoint bestanden!
>>> Zie http://www.gnu.org/philosophy/no-word-attachments.nl.html
>>> ------------------------------------------------------------------
>>>
>>>
>>> _______________________________________________
>>> genabel-devel mailing list
>>> genabel-devel at lists.r-forge.r-project.org
>>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/genabel-devel
>> _______________________________________________
>> genabel-devel mailing list
>> genabel-devel at lists.r-forge.r-project.org
>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/genabel-devel
>>
> 
> 
> 
> 
> _______________________________________________
> genabel-devel mailing list
> genabel-devel at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/genabel-devel
> 

-- 
-----------------------------------------------------------------
L.C. Karssen
Utrecht
The Netherlands

lennart at karssen.org
http://blog.karssen.org

Stuur mij aub geen Word of Powerpoint bestanden!
Zie http://www.gnu.org/philosophy/no-word-attachments.nl.html
------------------------------------------------------------------

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: OpenPGP digital signature
URL: <http://lists.r-forge.r-project.org/pipermail/genabel-devel/attachments/20121004/ed18c9e3/attachment.sig>