sd vs variances

[stephens999]

The output resid.sd appears to be the variance, not the sd.
The docs say it is the variance, so appear to be correct, but the name of the variable is very confusing.

[stephens999]

the initialization of resid.sd in init_mr_ash uses the sd though
resid.sd <- sqrt((n-1)/n) * sd(y - X %*% b)

[stephens999]

actually the prior sd may also be the prior variances!

Generally, mr.ash.alpha used variances (eg sigma2, sa2) so since we are interfacing
with that code we should probably stick with variances throughout to avoid errors.

[pcarbo]

Actually I think we were trying to make the interface more consistent with ashr; ash() has input argument mixsd.