I've done my best to root out all edge cases relating to n.sv being 1 or 0. For 1 SV, all sva variants should now return a 1-column matrix of surrogate variables rather than a numeric vector. For 0 SV, they should return a 0-column matrix of surrogate variables, a pprob.gam of all zeros (except for ssva, where pprob.gam = controls), a pprob.B of NULL, and n.sv = 0.
A test should probably be added running all the sva variants with 0, 1, and 2 or more SVs and assertting that the SVs are always returned in a matrix.
jtleek
commented
7 years ago
I've done my best to root out all edge cases relating to n.sv being 1 or 0. For 1 SV, all sva variants should now return a 1-column matrix of surrogate variables rather than a numeric vector. For 0 SV, they should return a 0-column matrix of surrogate variables, a pprob.gam of all zeros (except for ssva, where pprob.gam = controls), a pprob.B of NULL, and n.sv = 0.
A test should probably be added running all the sva variants with 0, 1, and 2 or more SVs and assertting that the SVs are always returned in a matrix.