Discrepency between R and C implementations

[TimTaylor]

Assuming the c++ implementation is correct then

finalsize/R/iterative_solver.R

Line 41 in 41ef432

contact_matrix_ <- contact_matrix * demography_vector %o% susceptibility

and

finalsize/R/newton_solver.R

Line 35 in 41ef432

contact_matrix_ <- contact_matrix * demography_vector %o% susceptibility

should be changed to

contact_matrix_ <- contact_matrix * susceptibility %o% demography_vector

or the micro optimized

contact_matrix_ <- contact_matrix * tcrossprod(susceptibility, demography_vector)

If/when the suggestion in #68 is taken up this will only need correcting in one place.
Tests that compare the c++ and R results to desired tolerance should be sufficient to test this.

[pratikunterwegs]

Thanks, I would say that this is fixed by this commit: 0019285, by preparing the contact matrix, after spreading through epi_spread, as so: t(t(contact_matrix * susceptibility) * demography_vector). Will close this issue along with the PR.

[Bisaloo]

Related: #7