Gauss-Hermite integration points
Calculate Gaussian Quadrature points for the Normal distribution using the abscissas and weights for Hermite integration.
theoretical number of quadrature points.
locations with weights that are less than this value will be omitted.
The conversion of the locations and weights is given in Lindsey (1992,
page 169:3) and Skrondal & Rabe-Hesketh (2004, page 165:1).
The argument numnodes is the theoretical number of quadrature points,
locations with weights that are less than the argument
be omitted. The default value of
minweight=0.000001 returns 14 masspoints
for the default
numnodes=20 as in Aitkin, Francis & Hinde (2005).
A list with two vectors:
locations of mass points
Nick Sofroniou (2005)
Aitkin, M., Francis, B. and Hinde, J. (2005). Statistical Modelling in GLIM 4. Second Edition, Oxford Statistical Science Series, Oxford, UK.
Lindsey, J. K. (1992). The Analysis of Stochastic Processes using GLIM. Berlin: Springer-Verlag.
Skrondal, A. and Rabe-Hesketh, S. (2004). Generalized latent variable modelling. Boca Raton: Chapman and Hall/CRC.
1 2 3 4 5
gqz(20, minweight=1e-14) # gives k=20 GH integration points. These are used in alldist # and allvc as fixed mass point locations when working with # option random.distribution='gq', and serve as EM starting points # otherwise.