I have some changes ready to commit that will implement qnorm() and the log-gamma function, lgamma(). I consider these two functions to be a relatively high priority in this wishlist, because they will enable a lot more custom-written likelihood functions to be written as mxAlgebras.
I can probably churn out trunc() and round() soon, too.
#1
I have some changes ready to commit that will implement
qnorm()and the log-gamma function,lgamma(). I consider these two functions to be a relatively high priority in this wishlist, because they will enable a lot more custom-written likelihood functions to be written as mxAlgebras.I can probably churn out
trunc()andround()soon, too.Log in or register to post comments
#2
Also, note that we already have
pnorm(). It's calledomxMnor().Log in or register to post comments
#3
I agree with your prioritization. Is your qnorm mxAlgebra also multivariate?
The mnemonic for omxMnor is Multivariate NORmal integration.
Log in or register to post comments
#4
Nope, it's univariate only. Quantiles from multivariate distributions are not necessarily unique. For instance, load the mvtnorm package and try
pmvnorm(lower = c(-Inf,-Inf), upper = c(1,0), mean = c(0,0), sigma = matrix(c(1,0.5,0.5,1),nrow=2))
pmvnorm(lower = c(-Inf,-Inf), upper = c(0,1), mean = c(0,0), sigma = matrix(c(1,0.5,0.5,1),nrow=2))
You'll see that (0,1) and (1,0) could both be considered the 0.468743 quantile of this distribution.
Log in or register to post comments
#5
Of course! I believe I was thinking of dnorm. But we already have that as the fitfunction of a Normal Expectation.
Log in or register to post comments