PLINK now takes imputation data for analysis by using --dosage. I suppose it automatically runs additive model. I would like to analyze data with recessive model and have tried to put --recessive in the commands. But the result turns out to be the same. So I would like to know how to correctly run recessive model for imputation data. thank you.

the dosage data makes it nearly impossible to test additive of dominant model. I guess you could put a threshold on the allele dosage and classify as homozyguous or not, but then you need to go out of plink - possible in R.

If you have genotype probabilities, not dosage - as most of the imputation programs can give if you ask them, then SNPTEST could run a recessive model by running a logistic regression where the likelihood is weighted by the genotype probability. In my experience, you need quite a lot of individuals to have enough rare homozyguous and get results - and the SNP needs to be frequent.

So anyways, to help you, we need some info about your input data.

The very nature of using imputed data, which are a form of dosage effects correlates with an additive model, dictates that testing with a recessive model is not possible. In order to run the recessive model, it is necessary to use only the original genotypes and no imputed data.

Once you have preliminary associations from the original genotypes, you can consider a type of fine mapping to further define the SNPs in a recessive model. This is not simple as the imputed must be treated by setting a threshold (which I myself do not understand) before you can combine with the original genotypes. Nonetheless, this is possible.