Dec. 31, 2021. Version 1.32
1. Improved function el2.test.wtm() by eliminating two
for-loops. It is now faster.
For larger samples, more improvement in computing time.
For example: for sample sizes (72, 41), [the aSAH data set]
the computing time is less than 1/5 of the old version.
(5.5 times faster).
2. Also in the el2.test.wtm(), change b%*%solve(A) to solve(A, b)
3. Modified el2.test.tws(). We added a new input: lamOld.
This can be used to improve computation of root (lambda)
in iterations.
You do not have to always start from 0 to search for root,
you may start from the previous solution.
Although the default is lamOld=0.
4. Some minor modification on the determination of interval
[lo, up] for the uniroot() to find root (lam).
Aug. 31, 2024 Version 1.33
Due to faster computers now, increase the default maxit value:
1. Increase the DEFAULT maxit=25 to maxit=50 inside the function el2.cen.EMs()
2. Increase the DEFAULT maxit=15 to maxit=35 inside the function el2.cen.EMm()
3. Increase the DEFAULT maxit inside the function el2.test.wtm to 35.
4. Add tol.u and tol.v to function el2.cen.EMs()
which is the L1 norm of probability vector change after each iteration.
Default to tol.u=tol.v=1e-6.