spatcounts (version 1.1)

Clarketest: Clarke's test for non-nested model comparison

Description

'Clarketest' suggests the better of two (not necessarily nested) models according to Clarke's statistic for the parameters in each of the iterations.

Usage

Clarketest(LogLike1, LogLike2, alpha = 0.05, p = NULL, q = NULL, 
correction = TRUE)

Arguments

LogLike1, LogLike2

the output of two model fits obtained by using 'LogLike'.

alpha

significance level, defaults to 0.05.

p, q

the number of estimated coefficients in models LogLike1 and Loglike2, respectively.

correction

boolean, if TRUE (default), the Schwarz correction will be used on the differences of log-likelihoods.

References

Clarke, Kevin A. (2003). Nonparametric Model Discrimination in International Relations. Journal of Conflict Resolution 47(1), 72-93.

Schwarz, G. (1978). Estimating the Dimension of a Model. Annals of Statistics 6, 461-464.

See Also

Vuongtest

Examples

data(sim.Yin)
data(sim.fm.X)
data(sim.region)
data(sim.gmat)
data(sim.nmat)

poi <- est.sc(sim.Yin, ~1+sim.fm.X[,1]+sim.fm.X[,2], sim.region, 
model="Poi", sim.gmat, sim.nmat, 3)
nb <- est.sc(sim.Yin, ~1+sim.fm.X[,1]+sim.fm.X[,2], sim.region, 
model="NB", sim.gmat, sim.nmat, 3)

DIC.poi <- DIC(sim.Yin, ~1+sim.fm.X[,1]+sim.fm.X[,2], sim.region, poi)
DIC.nb <- DIC(sim.Yin, ~1+sim.fm.X[,1]+sim.fm.X[,2], sim.region, nb)

ll.poi <- LogLike(sim.Yin, ~1+sim.fm.X[,1]+sim.fm.X[,2], sim.region, poi)
ll.nb <- LogLike(sim.Yin, ~1+sim.fm.X[,1]+sim.fm.X[,2], sim.region, nb)

Clarke.poi.nb <- Clarketest(ll.poi, ll.nb, alpha = 0.05, p = DIC.poi$p.D, 
q = DIC.nb$p.D, correction = TRUE)