| Library: | metrics |
| See also: | hhtest glmmultlo genmultlo |
| Quantlet: | hhmult | |
| Description: | hhmult calculates the H-H statistic to jointly test the specification of the link functions of a polychotomous response model (such as the conditional logit model or the multinomial logit model) |
| Usage: | {t,p} = hhmult (vhat,y,yhat,h) | |
| Input: | ||
| vhat | n x M matrix containing the projected indices | |
| y | n x M matrix containing n observ. of M binary dependent variables | |
| yhat | n x M matrix with the predicted probabilities P(Y_m=1|x) | |
| h | 1 x M vector of bandwidths | |
| Output: | ||
| t | scalar, the value of the test statistic | |
| p | scalar, the p-value of t, i.e. 1 - cdfc(t,M) | |
library("glm")
library("metrics")
n = 100
b =(1|2)~(-1|0)~(1|1)
x = normal(n,rows(b))
y = genmultlo(x,b)
ml = glmmultlo(x,y)
vhat= x*ml.b
y1 = y.=1
y2 = y.=2
y3 = y.=3
y = y1~y2~y3
c1 = cols(vhat)
h = matrix(c1)'
yhat= ml.mu
{t,p}= hhmult(vhat,y,yhat,h)
t~p
test statistic, t, and p-value, p.