Hello
I am using SNPTEST v2.5.2 on a Linux server. I am running a GWAS on different samples from different ethnic groups and for some of them I am not obtaining any association results. I obtain the error "model_not_fit:design_matrix_singular_value_below_limit" for all SNPs.
My command is:
./snptest_v2.5.2 \
-data vcf_file sample_file\
-o output_file \
-frequentist 1 \
-method expected \
-cov_names nn_gender c_id V1 V2 V3 V4 V5 V6 V7 V8 V9 V10 \
-pheno z_logcpep \
-genotype_field GP
The phenotype is a continuous trait and the covariables are sex (nn_gender, discrete, coded 1/2), participant centre ID (c_id, discrete, up to 10, coded as letters) and PCs 1-10 (V1:V10).
I think I have traced down the issue to the centre ID covariable since it runs when I exclude it from the model. I have tried re-coding it as integers, as separate dummy variables, and removing the last centre (so including only up to centres N-1), but these steps have not helped.
Interestingly I found that if I excluded participants from a centre where there were fewer than 10 participants in the sample it did run. I don't mind necessarily excluding these participants since the overall number is low, but I was wondering if anyone could explain why this might be happening? What I don't understand is that I have a few samples with participants from a centre n<10 and it will run fine. Also, in samples where I do exclude participants, the centre that ends up not being in the model is not the same across samples because they have varying n. I would appreciate some input to make sure it is not an error with the data or on my part.
Many thanks
Alice
To unsubscribe from the list visit this webpage https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=OXSTATGEN&A=1
|