3

I apologize if this question is dumb as all get out.

I want to leverage R's mle2() function to find optimum parameters to a particular statistical function; I presume it does so using gradient descent? So I've got my call like this: 

r = mle2(minuslogl = likelihood, 
    start = list(a1=0.1,b1=0.1,x01=0.1,d2=0.1,b2=0.1,x02=0.1,c=1), 
    data = list(values=v,data=d))

Where my likelihood function specifically needs a1, b1, d2 and b2 to be within the range [0,1] (the real numbers between 0 and 1). If mle2() uses gradient descent, I'm assuming it starts moving the mentioned parameters into the negative range during its optimization phase - but I want it to specifically not do that, I want it to search the parameters between 0 and 1.

Is there a way? Am I being really ignorant here?

Thanks in advance.

1 Answers1

3

I found the answer. There is documentation for what I want; though I wasn't able to find what I wanted immediately in the pages I was viewing.

There are literally lower and upper parameters to the mle2() function.

Examples can be found here:

http://www.inside-r.org/packages/cran/bbmle/docs/mle2

Here is a functioning example of what to do: 

r = mle2(minuslogl = likelihood, 
     start = list(x01=0.1,x02=0.1, c=1, a1=.1, b1=.1, d2=.1, b2=.1), 
     data = list(values=v,data=d),
     lower = c(a1=0,b1=0,d2=0,b2=0),
     upper = c(a1=1,b1=1,d2=1,b2=1), 
     method="L-BFGS-B")

This bounds the a1, b1, d2, and b2 variables and also gives them starting values for the gradient descent method.