regarding "Model estimation FAILED! Returning starting values." warning

[dos7...@gmail.com]

Hi gSEM team,

When I run userGWAS with fix_measurement = TRUE and provide multiple SNPs at once as input (i.e., SNPs = output_of_sumstats), I sometimes get the message:

“Model estimation FAILED! Returning starting values.”

However, when I run the analysis one SNP at a time 

(i.e., SNPs = output_of_sumstats[1, ]), I often do not see this warning.

After reviewing the code, my current hypothesis is that when multiple SNPs are analyzed in a single run, the base model remains fixed across iterations, and that this could be related to why the warning appears in some cases.

Also, I confirmed with the lavaan author that the message “Model estimation FAILED! Returning starting values.” may be printed due to convergence issues, where starting values are not actually returned (see: https://github.com/yrosseel/lavaan/issues/448).

Given this, would it make sense to run userGWAS using the full summary statistics first, and then (1) refit only the subset of SNPs that produce warnings, or alternatively (2) fit each SNP individually? 

Thank you.