Hi Giovanni,

You can try -method newml which generally attempts to fit every SNP it can - this may recover some SNPs.

However, you should treat the obtained P-values with caution because they rely on an asymptotic approximation which may not hold well when the frequency is low.  (If you see a big discrepancy between the SNPTEST P-value and a Wald test P-value computed from the beta and standard error, that's an indication that this is occuring.)

Best wishes,
Gavin.

On 6 Nov 2013, at 11:18, giovanni acierno wrote:

Dear Gavin,
the problem is that we are interested in those snps with very low maf. Checking the output I obtained by using snptest, snps with lower maf with respect to all of the other are confirmed in dbsnp as variants with low maf. How can I bypass the problem? Does snptest consider a threshold which I could modify?
Exist other similar software on the web you could suggest me?
Many thanks 
Giovanni



2013/11/6 Gavin Band <[log in to unmask]>
HI Giovanni,

I think your results are as expected - SNPTEST is unable to fit the model at very rare SNPs.

Thanks,
Gavin.

On 6 Nov 2013, at 09:40, giovanni acierno wrote:

Dear Gavin,
attached is the log file deling with my last launch of snptest_v2.5-beta4.
I attached here below some lines of  my output file.
Many thanks 
Giovanni


2013/11/5 Gavin Band <[log in to unmask]>
Hi Giovanni,
Re: -total_prob_limit: are you using the latest version (2.5-beta4)?  If not please try that version.
Please send log files / some lines of output if you want help diagnosing a problem.
Thanks,
Gavin.

On 5 Nov 2013, at 15:08, giovanni acierno wrote:

Dear colleagues,
I'm using the second version (V2) of Snptest software. Starting from imputed snps obtained from Impute2 analysis (80.000 imputed snps starting from 1071), I have changed the -total_prob_limit parameter in a range from 0.001 to 1 paired with -method score and -method em respectively. No differences in the output files sum up. Impute2 output corresponds to region 2q35 using hg19 human genome. Out of about 80.000 snps ranging in that region, more than a half have not a related p-value. Is in your experience a recurrent issue? Is the fraction of discarded snps too high?
Many thanks, 
To unsubscribe from the list visit this webpage https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=OXSTATGEN&A=1

--------------------------------------------------------------

Dr. Gavin Band

Wellcome Trust Centre for Human Genetics
University of Oxford


--------------------------------------------------------------







To unsubscribe from the list visit this webpage https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=OXSTATGEN&A=1

To unsubscribe from the list visit this webpage https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=OXSTATGEN&A=1 <log><output.xlsx>

--------------------------------------------------------------

Dr. Gavin Band

Wellcome Trust Centre for Human Genetics
University of Oxford


--------------------------------------------------------------







To unsubscribe from the list visit this webpage https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=OXSTATGEN&A=1

To unsubscribe from the list visit this webpage https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=OXSTATGEN&A=1

--------------------------------------------------------------

Dr. Gavin Band

Wellcome Trust Centre for Human Genetics
University of Oxford

+44 1865 287532

--------------------------------------------------------------







To unsubscribe from the list visit this webpage https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=OXSTATGEN&A=1