why nbeta works for 4 or more inputs?

[chy06]

Hi. I've read the closed issue about fact that nbeta has 3 parameters. But as I tried for example, nbeta(c(0.3,0.1,0.2,0.4),2), or using midas_r with a starting value of length 4 or more, it still gives me something. I wonder what it means.

[vzemlys]

It is all in the code:

> nbeta
function (p, d, m) 
{
    nbetaMT(c(p, 0), d, m)
}

> nbetaMT
function (p, d, m) 
{
    eps <- .Machine$double.eps
    xi <- (1:d - 1)/(d - 1)
    xi[1] <- xi[1] + eps
    xi[d] <- xi[d] - eps
    nb <- xi^(p[2] - 1) * (1 - xi)^(p[3] - 1)
    if (sum(nb) < eps) {
        if (abs(p[4]) < eps) 
            rep(0, d)
        else p[1] * rep(1/d, length(nb))
    }
    else {
        w <- (nb/sum(nb) + p[4])
        p[1] * w/sum(w)
    }
}

Function nbeta calls nbetaMT, which was taken from MATLAB MIDAS toolbox. The difference from the usual normalized beta is that additional constant is added after normalization. This constant is passed as a fourth element of the parameter vector. If it is zero then nbetaMT and nbeta give the same result. All the other elements of the parameter vector are ignored.

[chy06]

I see, thank you!

[AdamElderfi]

Hi - I have a similar question. What is the fourth parameter achieving? I understand that we have a normalising constant and two hyperparamters - but I can't see the purpose of the fourth parameter?

[vzemlys]

When I was developing the midasr package, I tried to make it compatible with MIDAS toolbox for Matlab. The Matlab toolbox had the fourth parameter. As you can see from the code it acts as a sort of bias, i.e it shifts the coefficients.