uniroot error

[sashagusev]

Hi, I'm getting an error when running SPACox on a matrix of 50k SNPs that looks like this:

Error in uniroot(K1_adj, c(-20, 20), extendInt = "upX", G2NB = G2NB, G2NA = G2NA,  : 
  no sign change found in 1000 iterations

I'm guessing this is a convergence issue from the uniroot call in GetProb_SPA. There's nothing unusual about this variant being tested (MAF or distribution) so I suspect this is just something sporadic about the optimization. However, the entire run crashes as a result and this happens in most of the batches, making it impossible to finish the analysis.

Is there some way a try/catch can be added around this call and NA's returned if the optimization fails?

[sashagusev]

In case it's useful, I've forked a quick hack to fix this here:
https://github.com/sashagusev/SPACox/blob/master/R/Library.r
This adds a try around the entire uniroot call and then returns p-values of 2 in instances where the uniroot did not converge. Ideally these should be returned as NAs but that broke some downstream processing.