I am running single variant tests (sample size ~60K) using SAIGE v1.1.6 and I've been attempting to utilize the sparse GRM to reduce computation costs. This has worked great for step 1, however, I am unsure about if/when the sparse GRM should be included in step 2 for single variant tests. I noticed on the wiki that you state: "If the sample size is relatively small, e.g. N < 1000, may consider using sparse GRM in step 2 using --sparseGRMFile, --sparseGRMSampleIDFile, --is_fastTest=TRUE", however, the step 2 page for single variant tests also indicates that you should either specify (or not specify) the sparse GRM files in step 2 depending on the version of SAIGE you are using and whether or not you estimated variance ratios in step 1. Is there a benefit of specifying the sparse GRM files in step 2 if the variance ratios are estimated in step 1?
Any insight you have on this is much appreciated! Thank you!
Emily
Hi Wei,
I am running single variant tests (sample size ~60K) using SAIGE v1.1.6 and I've been attempting to utilize the sparse GRM to reduce computation costs. This has worked great for step 1, however, I am unsure about if/when the sparse GRM should be included in step 2 for single variant tests. I noticed on the wiki that you state: "If the sample size is relatively small, e.g. N < 1000, may consider using sparse GRM in step 2 using --sparseGRMFile, --sparseGRMSampleIDFile, --is_fastTest=TRUE", however, the step 2 page for single variant tests also indicates that you should either specify (or not specify) the sparse GRM files in step 2 depending on the version of SAIGE you are using and whether or not you estimated variance ratios in step 1. Is there a benefit of specifying the sparse GRM files in step 2 if the variance ratios are estimated in step 1?
Any insight you have on this is much appreciated! Thank you! Emily