Testing with covariates got errors "gsl: brent.c:202: ERROR: function value is not finite" about gemma HOT 3 CLOSED

@Owen-haha I would encourage you to download the latest version of GEMMA. Please let me know if you have trouble getting it to work.

Second, I would recommend starting with a simpler association analysis such as this:

./gemma -g gen_file.txt -p pheno_file.txt -lm 2 -maf 0

You will find that the results aren't very interesting for your particular example because there is very little variation in the SNP you are testing; the minor allele is observed only once among the ~1,500 samples. Regression coefficient estimates in this sort of case aren't well behaved. I would suggest that you remove SNPs such as this that exhibit very little variation in your sample.

from gemma.

Hi @Owen-haha, can you please give the exact gemma command you used? Do you get the same error if you do not include the polygenic term u (that is, you use the -lm option)? Are SNP1 –4 also being tested, or did you remove them from the genotype data file?

from gemma.

Hi @pcarbo , below is the command i run. I also attached a real sample here.
gemma.0.93b -g gen_file.txt -p pheno_file.txt -a snp_info_file.txt -k realmatrix_file.txt -lmm 4 -o prefix -maf 0 -c cov_file.txt
gen_file.TXT
pheno_file.txt
snp_info_file.txt
realmatrix_file.txt.zip
cov_file.txt
I cannot do -lm because the version i used dose not support it.
There is only one SNP in gen_file.txt, and 9 covariates in cov_file.txt. That means each time only tests one SNP. And the SNPs in gen_file.txt and cov_file.txt are excluded.

from gemma.